biometric technology research papers

Advertisement

Biometric applications in education

• Technical Paper
• Published: 28 July 2021
• Volume 15 , pages 365–380, ( 2021 )

Cite this article

• Marcela Hernandez-de-Menendez 1 ,
• Ruben Morales-Menendez   ORCID: orcid.org/0000-0003-0498-1566 1 ,
• Carlos A. Escobar 2 &
• Jorge Arinez 2  

15k Accesses

18 Citations

80 Altmetric

10 Mentions

Explore all metrics

Educational institutions are acquiring novel technologies to help make their processes more efficient and services more attractive for both students and faculty. Biometric technology is one such example that has been implemented in educational institutions with excellent results. In addition to identifying students, access control, and personal data management, it has critical applications to improve the academic domain's teaching/learning processes. Identity management system, class attendance, e-evaluation, security, student motivations, and learning analytics are areas in which biometric technology is most heavily employed. A literature review is performed to present an overview of biometric technology applications for educational purposes, challenges that must overcome to implement biometric technology, and potentially foreshadowing trends effectively. The future seems promising for biometric technology; the biometric technology market is expected to reach a value of USD 94 billion by 2025 at a compound annual growth rate of 36%. New characteristics are under development for commercial applications, such as vascular pattern recognition, ear shape recognition, facial thermography, odor sensing, gait recognition, heartbeat authentication, brain waves, and human body bioacoustics. The biggest challenge this technology must overcome is security and privacy issues, which must be addressed to fully develop the technology to its full potential. It is desirable that this literature review can provide researchers with a sound vision of the potential that biometric technology will have in education.

Similar content being viewed by others

Biometric Tools for Learner Identity in e-Assessment

Using Biometrics to Support Affective eLearning for Users with Special Needs

Ensuring Online Exam Integrity Through Continuous Biometric Authentication

Avoid common mistakes on your manuscript.

1 Introduction

Educational institutions are transforming themselves to be more flexible and adaptable to respond to actual social demands. They are acquiring novel technologies to help them make processes more efficient and services more attractive for students and faculty (and society). Competency-based teaching is oriented to progressive and gradual learning of knowledge, skills, attitudes, and values, which requires real experiences in the educational process.

Generational changes have rendered it necessary for universities to understand and adapt to student's requirements and expectations. Generation Z (GenZ), born in 1995, grew up during the technological revolution (internet, social networks, etc.). Thus, they are easily linked to the virtual world [ 14 ]. GenZ has some common characteristics: creativity, self-taught, multi-screen, and multi-tasking. Therefore, they represent a challenge for traditional education:

47% of GenZ spends more than 3 h a day on a video platform.

59% uses YouTube to learn.

39% prefer to attend a class with a teacher [ 52 ].

Their pragmatism distinguishes GenZ, adaptability to work conditions, and integration with other generations (Baby Boomers, GenX, and Millennials). This generation brings new challenges for education, new uses of technology (and new technology), new ways of teaching and evaluating, and new digital platforms. These new educational tools and approaches have only been accelerated in response to the Covid-19 pandemic.

Educational institutions have implemented various technological tools, including virtual and augment reality, the internet of things, artificial intelligence, 3D printing, telepresence, blockchain, and biometrics [ 28 ]. However, the last one is gaining the attention of the educational community as biometric identification systems are becoming popular.

Biometric “is the science of analyzing physical or behavioral characteristics specific to each individual to authenticate their identity.” In a few words, it measures the human body. This technology can measure physiological (e.g., fingerprints, vein patterns, iris, retina, the shape of the hand and shape of the hand) and behavioral traits (e.g., voice recognition, gait, gestures, the sound of steps, and signature). A combination of characteristics can also be made as a multimodal biometric, which improves measurement confidence. Such combinations could be face and fingerprint, face and iris, etc. The main benefit of biometric technology is that it collects unique human characteristics for each person [ 64 ].

Biometric systems have characteristics that allow for their use in different domains. Several government initiatives (e-passports, e-driver's licenses, border management, and national identifications) are being implemented using advanced biometrics. The banking sector has improved security by using biometric technologies (fingerprint scanners, facial/voice recognition, etc.) to track its customers. Banks also use biometrics to authenticate and protect customer's financial data; it is a long-term market change. With the Internet of Things (IoT), collecting and storing patient data in hospitals has become easier/safer; however, biometric technology can significantly improve. Biometrics can improve essential aspects such as patient's identity and support with real-time information to offer adequate medical services. Facial recognition, iris, and finger scanners are examples of biometric technologies used to access or control a patient's data.

Regarding the security and privacy of patient data, biometrics helps in monitoring and documentation. The productivity of medical personnel (nurses, doctors, etc.) can also be monitored using biometrics. The hotel industry is adopting this technology to achieve a strategic differential in this highly competitive sector. This technology (fingerprints or facial/voice recognition) allows hotels to efficiently automate guest's identification and access and make their stay more comfortable and safe. Using biometric technology, the entry in the migration processes in airports can be accelerated while maintaining security. Many systems use fingerprint scanners, a very efficient and safe tool. Many retailers rely on biometric technology (facial recognition) to prevent theft or promote customer-related items based on their views and interests in a product (Sharman, n.d.). There are platforms such as elBio that are built to teach students about biometric technology. This system introduces the field through a tutorial, and real interactions with verification and identification processes are performed [ 59 ]. Hand geometry systems are helpful to allow access to university systems [ 51 ].

Biometrics is not a new technology. This science dates back to ancient Egypt, where pharaohs relied on height measurements to verify their identity. They also authenticated the decrees they issued by adding their fingerprint to a document and their signature [ 4 ]. Figure  1 indicates the criteria that must be met for a physiological or behavioral attribute to be used as a biometric trait [ 33 ].

Criteria of a biometric trait

A classification of the characteristics measured with biometric systems is presented in Table 1 [ 68 ]. In addition, feature sets are included between brakes [].

Biometric technology is also used for recognizing individual human identities, including DNA, retinal, iris, voice, facial images, fingerprints, and handprints. It scans one of the above characteristics and compares them with stored data. The most common features used are fingerprints and iris. Fingerprints are likely a standard method because fingerprint scanners are relatively inexpensive [ 40 ]. Fingerprint scanners are applied in everyday activities, such as access control and vehicle driver license registration. Research in iris biometrics has accelerated since 2001, with its peak occurring in 2006.

Features such as the face, voice, and retina are also employed. Face biometrics is fast, inexpensive, and non-intrusive. It is also a robust technology that is not heavily influenced by the environment. Researchers have also been working on voice systems to understand the personal encoding of natural languages. In addition, voice biometrics can work with standard telephones, making it a versatile option [ 77 ].

Behavioral biometrics provide several advantages over traditional biometric technologies (iris, fingerprint, handprints, etc.). They can be collected non-obtrusively or even without the knowledge of the user. The collection of behavioral data often does not require any specialized hardware and is very cost-effective. While most behavioral biometrics are not unique enough to provide reliable human identification, they have been shown to give sufficiently high accuracy identity verification. Unfortunately, this type of biometric is not as popular as traditional biometric systems. The reason could be that behavioral characteristics are not exact enough to identify people. However, they have many advantages, such as analyzing people in their routine activities and collecting information that includes a time dimension. The behavior is quantified, and results are used to verify the identity of the person. Some behavioral biometric examples include facial features, email behavior, game strategy, lip movement, mouse dynamics, painting style, programming style, signature/handwriting, storage activity, and systems calls [ 75 ]. Biometrics can also be used for verification or identification processes. In the former, a person that claims identity is verified or rejected. An unknown sample is matched with the stored data [ 5 ].

A basic biometric system is a pattern recognition system that matches the acquired image's salient or discriminatory features with stored image's features. It comprises five main modules, contained in Fig.  2 [ 19 , 68 ].

Composition of a biometric system

A biometric system works in verification or identification mode. In the verification mode (or positive recognition), a user submits the biometric trait and particular identity to the confirmation system. The recognition system validates or voids the user’s claim by comparing (one-to-one) the proposed biometric feature and stored biometric feature associated with the identity. In identification mode (or negative recognition), the system tries to recognize the user by comparing the submitted biometric feature to all the stored biometric features in the database by making (one-to-many) comparisons without a specific identity claim.

The recognition systems based on the measurement of a single biometric modality cannot guarantee 100% accuracy, as shown in Table 2 [ 68 ]. This is due to several factors: (1) noisy data, (2) class variation, (3) distinctiveness, (4) non-universality, and (5) false identity. One solution to this challenge is to use multiple and independent biometric modalities that offer more robustness in precision. A multimodal biometric system may involve multi-sensors, multi-algorithms, multi-instances, multi-sample, multimodal systems. This approach could be serial or cascade mode and with different fusion levels: sensor, feature, matching score, or decision.

Multimodal processes are done through different devices/sensors that can be used to collect biometric data; these include: (a) cameras for facial or iris recognition, (b) scanning devices for fingerprint recognition, (c) smartphones and laptops for fingerprint recognition, (d) telephones and microphones for voice recognition, (e) special tablets and a digital pen for signature recognition, and (f) keyboard device for keystroke dynamics [ 15 , 16 ].

The biometric market is expected to reach a value of USD 94 billion by 2025, at a compound annual growth rate of 36%, during the next 5 years. This is due to a paradigm shift in commercial discourse, seeking new comprehensive solutions, more privacy, and fewer security threats. Vendors of biometric solutions include: JayPeetek Inc., SecuGen ® Biometric Solutions, onClick® Corp, etc. which offer biometric mice [ 40 ], Gladstone Education offers identity management systems; Squidcard provides a cashless payment system using fingerprint; Voice Commerce offers voice authentication services [ 23 ], and Newcastle-based ievo Ltd is a producer of advanced biometric fingerprint readers who is very demanded by schools and also nurseries [ 18 ]. The benefits of using biometric systems vary, including increased security, increased convenience, reduction of fraud, and delivery of enhanced services [ 77 ].

A literature review is performed to present an overview of biometric technology applications for educational purposes. The paper outline is as follows: Sect.  2 describes the applications biometrics offers in the academic field. Section  3 presents the challenges that must be overcome to implement the technology under consideration effectively. Section  4 presents the trends that are foreseeing. Finally, Sect.  5 concludes the paper.

2 Applications in education

Biometrics technology is gaining momentum. In 2001 the MIT Technology Review considered biometrics a world-changing emerging technology [ 59 ]. Biometric technologies are disrupting several industries and sectors. General applications include its use for recreational activities, such as in Disneyland. It can also be used for replacing password systems. Innovative devices have also been developed, such as a mouse that recognizes the fingerprint of its owner. ATMs also use this technology [ 40 ].

There is also another exciting field of biometrics, in which the emotional and cognitive state of people are detected. This can be used to monitor student's behavior/emotions and change the educational process appropriately. For example, researchers argue that boredom negatively influences learning, whereas engagement improves learning outcomes; biometric sensors have been used to measure electrodermal activity, skin temperature, and heart rate, all good predictors of emotions [ 73 ]. Biometrics allows academic institutions to save time, money and also improve educational and non-educational activities. They also offer convenience, safety, and security. Various applications are identified: school access, control of attendance, food service, access to library and media center, bus transportation, control staff time, among others (Fry and Dunphy, n.d.). In addition to identifying students, access control, and personal data management, it has critical applications to improve teaching/learning processes in the educational domain, Fig.  3 .

Biometrics in the educational domains

2.1 Identity management system

Biometric person recognition systems share many issues and challenges with other pattern recognition applications like video surveillance, speech technologies, human–computer interaction, data analytics applications, behavioral modeling, or recommender systems. Identity Management Systems ( IMS ) are platforms where a permission device gives access to a specific service. These are used in education to provide access to a given product or service or for electronic registration. Passwords are very unsafe and sometimes created to be easy to remember. Biometrics are used in this area for security reasons, as it can grant access to a given system to only authorized persons, identified by their physical or behavioral characteristics [ 59 ]. Fingerprint cards are already used in schools for students that acquire free meals in coffee shops [ 23 ]. Indeed, the use of fingerprints is a common practice from elementary school to universities/research centers.

Nita and Mihailescu [ 48 ] proposed a secure e-learning system based on biometric authentication and homomorphic encryption exploiting cloud computing. Additionally, this proposal predicts if he/she would pass a final exam based on past data of the user’s behavior (using biometrics data).

2.2 Class attendance

With biometrics, the attendance of students to a given class can be accelerated. This is advantageous since time devoted to taking attendance is reduced. Also, a more accurate registration process can be performed, diminishing errors [ 9 ]. This technology also enables identifying causes and patterns of absence, and the behavioral characteristics of students can be correlated with class achievement. Analysis of absence between year groups and groups of individuals can also be performed. Universities can also use this technology to track students [ 60 ].

The University of Sunderland London Campus is already using this technology to report class attendance. They use a portable device that has a fingerprint sensor. When students enter the class, they put their fingers on the device to easily register their presence. In India, Delhi University uses this system to track professors’ attendance to class [ 10 ]. In addition, researchers used a biometric fingerprint device to improve active class participation in those classes that consist of a pure lecture. It was combined with a rewarding activity resulting in improved student engagement and class results [ 32 ]. Finally, this technology can also be used in online activities or education to manage time effectively [ 29 ].

2.3 e-Evaluation

The submission of e-exams is a relatively new use of biometrics (i.e., since 2017). However, research in this area has been headed by international organizations such as the European Union with its project Adaptive trust-based E-system Assessment for Learning. In these projects, 17 European organizations use keystroke and facial recognition technologies to identify university student's identities and reduce cheating [ 25 ].

Fingerprint recognition can be used for students to take online exams. In this case, learners must verify their identity, and only after that, the exam can be shown on the screen. This can be done by identifying them through facial, iris, or voice recognition. Either the characteristic used, a student sample must be first taken to perform the matching process [ 15 , 16 ].

Biometric systems (finger scans) to verify the identity of IELTS test takers have been implemented globally. As a result, IELTS is available in over 900 locations in 130 countries, making the British Council and IDP IELTS [ 10 ].

2.4 Security

Biometrics is a useful technology to identify students and ensure no outsiders either in class or on the university’s campus. On the other hand, this technology can be used in combination with surveillance cameras to detect strangers. In addition, a blocking protocol can be activated in an emergency on campus to ensure that no one enters or leaves. Using identification chips (based on radio frequency) and combining them with intelligent data, students can be located to guarantee their safety. Biometric systems can also limit access to computers, emails, websites, and other restricted educational tools (assessments) [ 29 ]. Finally, students’ presence on campus can be tracked by knowing at any time where they are, when they arrived and left, and where they went [ 18 ].

Higher Education is becoming one of the most popular targets for cyberattacks because universities have relatively open networks. For example, universities have several wireless networks that connect their areas using multiple bandwidths; these multiple networks lead to an output that contains student data (payment information, social security number, personal addresses, etc.). Additionally, universities must comply with various laws to protect student data; these law's guidelines may restrict the institution's IT infrastructure or leave it vulnerable to hackers (Bio-KeyTM, n.d.).

The computer industry is changing rapidly, and people are pushing toward a new technological system buying the latest editions of phones and computers. However, inexperienced users coming into technology rapidly, many of them could unintentionally expose information without realizing it. An inexperienced user can potentially be subject to scams, spoofing, and phishing because the university network allows hackers to enter and exit a system without being detected swiftly. Educational institutions establish an open network architecture with multiple access points; if someone misplaced their cellular phone, a hacker could potentially log into the system and access the entire mainframe.

If intellectual property (patents, documented permission) is stolen, it could cost the institution a large amount of money. Faculty and student personal identification are available; it includes healthcare, credit card/payment, etc. Also, students and faculty have highly sensitive data (bank accounts, personal addresses, etc.). There are many government regulations that Higher Education institutions must follow. Still, in doing so, they expose themselves to a possible attack that they cannot stop due to the regulations. Cyber attacks in Higher Education date back to 2002, Table 3 [ 41 ]. The hackers’ goal remains personal data, social security numbers, financial information, opening up a new credit card, collecting tax refund, etc.

Biometrics enable body-based security—a technology that authenticates identity based on physical characteristics such as fingerprints, irises, facial structure, voice, and even gestures. Fingerprint authentication (the most commonly used biometric technology) is based on a unique set of identification. Instead of using a passcode, only one person can log into a system using this biometrics technology. Students can start using their fingerprints as their credentials for several reasons: (a) the authentication rarely fails, (b) fingerprints do not rely on memory, (c) rapid system recognition of the fingerprint. By establishing fingerprint authentication, biometrics can protect network architecture and access to other areas. Biometrics is going to be the most powerful technology against cyber-attacks in Higher Education.

2.5 Understand students’ motivation and academic progress

In addition to knowing the course's educational content, a good teacher must know their students very well and identify their cognitive status to guide the teaching–learning process properly. For example, if the teacher determines the student's commitment or motivation, he can use different educational strategies to optimize the teaching process. However, when the educational process is done remotely, and for many students, automated system's support is required. This is where biometric technology opens up an excellent opportunity to develop strategies that help detect student's cognitive states.

Due to their influence on learning, emotional states play a crucial role in education in general. Boredom has been shown to influence learning, while engagement can positively improve learning outcomes. Frustration and confusion can positively affect learning if the student can resolve these states. Estimating prediction in real-time of student's affective states is a research topic of great interest due to its benefits through different intervention strategies [ 24 ]. The collection of appropriate biometric data and the analysis of physiological and behavioral patterns during a learning experience can help introduce proper interventions to improve the learning experience as the main hypotheses in this domain.

Biometrics provides an objective measure of the physiological reactivity of users that is used to infer affective states. Electrodermal activity, skin temperature, and heart rate showed high performance as predictors of emotions [ 34 , 35 , 55 ]. Wampfler et al. [ 73 ] predict a student's affective states (while solving math exercises) using arbitrary writing and drawing assignments (based on stylus data).

A low-cost mobile setup to detect student's affective states (non-intrusive and minimum issues related to privacy) is proposed. The system considers bio-sensor data from skin conductance, heart measures, and skin temperature with handwriting data recorded by a stylus to predict student's affective state in a valence-arousal space of emotions proposed [ 54 ]. Valence describes how much emotion is perceived as positive/negative, and arousal represents the emotion's intensity. The circumplex model has two dimensions representing affective states in terms of valence/arousal. The circumplex model has the leading eight affective states (Arousal-0o, Excitement-45o, Pleasure-90o, Contentment-135o, Sleepiness-180o, Depression-135o, Misery-270o, and Distress-315o). The pleasantness-unpleasantness and arousal-sleep dimensions account for the significant proportion of variance; the dimensions of the effects are bipolar; any effect could be defined as combining pleasure and arousal components. Recorded stylus and bio-sensor data are preprocessed, and the relevant features are extracted to train a classification model (using the Random Forest algorithm) for the specific affective regions. Early results are very promising and practical; however, more experiments and validations are required to have overall effects in other knowledge and educational settings.

Dafoulas et al. [ 13 ] used a range of sensors measuring critical data from individual learners, including heartbeat, emotion detection (anger, disgust, fear, happiness, sadness, and surprise), sweat levels, voice fluctuations, and duration/pattern of contribution via voice recognition. Employing biometrics for supporting assessment, facilitating, and enhancing learning experiences in collaborative learning.

Smart biosensors and cameras (infrared) can identify and track students, diagnose their behavioral state (body language and eye contact) and their peer's actions. This can aid in making opportunely changes in the teaching–learning processes and improve students’ results. Also, online student engagement diagnosis will help teachers use the needed teaching strategies and technologies to optimize student's learning.

Many scientific projects have been developed to investigate and estimate student's cognitive states during the teaching/learning processes, based on data and evidence using biometric technology. With this information, the educational process can be optimized. Preliminary results are reasonable; however, research must continue to obtain general, practical, and valid conclusions in different domains. Based on biometric technology, it is possible to assess a student's academic progress and customize strategies to help him achieves his goals [ 29 ].

2.6 Biometrics in learning analytics

Learning Analytics is defined as measuring, collecting, analyzing, and reporting data about students and their learning contexts to understand and optimize the learning process and the learning environment [ 58 ]. Learning Analytics is an area of technology-enhanced learning, Big data, cloud technologies, virtual reality, brain-computer interface are some of the technologies that powered Learning Analytics .

LA benefits are: (1) prediction of student performance, (2) personalized student experience, (3) student confinement increases, (4) improves e-learning systems, (5) enhances cost-efficiency. To achieve these Learning Analytics benefits, it is necessary to collect data from various sources. Many data sources vary among many fields, including written or online surveys, interviews, students’ opinions, improvements suggestions, web tools, and more sensitive data: biometric data. The focus is on biometric data due to the recent technologies and devices that enable collecting and analyzing such data [ 15 , 16 ].

Social Network Analysis, GISMO (student-monitoring tool), CourseVis (learning management system), Contextualised Attention Metadata, LOCO-Analyst, Social Networks Adapting Pedagogical Practice ( SNAPP ), Honeycomb, Gephi, sense.us, Signals, and GRAPPLE Visualisation Infrastructure Service ( GVIS ) are some of the Learning Analytics tools that have been developed over time [ 20 ].

Secure access to information and data privacy of learners’ data shall be provided as one of the essential issues that biometric-based systems can quickly solve. Brain-Computer Interface ( BCI ) is a direct communication channel between the human brain and a computer. BCI is a Human–Computer Interface branch oriented towards research human cognitive, sensorial, or motor functions [ 38 ]. EEG -based BCI can help focus student attention and memory retention. It can also measure affective states and adapt the challenge difficulty to the learner’s emotions or even measure the engagement level. This area can improve learning and opens up a large area of research to help people with disabilities.

Over time, tracing Learning Analytics development highlights a gradual shift away from a technological focus towards an educational focus. Factors driving the growth of Learning Analytics :

Big data : Significant amounts of learner activity take place, and records are distributed across a variety of different sites with different standards, owners, and levels of access.

Online learning : Learning online offers many benefits, but it is also associated with problems.

3 Challenges

The growth of biometric systems has exponentially developed in education and other application areas; however, some opportunities and challenges must be considered to ensure good results.

3.1 Users reluctant to use the technology

As biometrics is relatively new, users might be reluctant to use it. It could pose various problems such as wide variance when measuring characteristics, affecting the system's performance. Cultural and social issues must be taken into consideration when designing a biometric system. If not, any initiative can fail. People could be reluctant and fearful of using these systems due since they collect personal information. Also, people may be afraid of acquiring a disease if using, e.g., a fingerprint apparatus (because of the Covid19 pandemic). Some persons might be fearful of how their photographs taken from recognition systems will be managed. Biometric systems that track employee's movement seem beneficial for the employer but intrusive for the worker [ 51 ].

3.2 People that can develop solutions and use the technology

For this technology to advance, it is necessary to educate future engineers in this technology well. This is important for developing commercial products that are attractive, useful, and at the same time, solve social problems. This could be achieved by incorporating knowledge regarding the technology in the curriculum of Information Technology universities programs. Several universities are already doing their part. For example, West Virginia University offers the first undergraduate program in biometrics; people who acquire a Bachelor of Science degree in Biometric Systems.

On the other hand, the University of Hertfordshire in the UK has already created a Master’s Degree course in Biometric and Cybersecurity. However, as this is a new field of study, it isn't easy to transmit knowledge effectively. Therefore, some researchers suggest the use of simulation tools.

3.3 Confiability

Biometric recognition is based on statistics; i.e., it does not perform a yes or no decision but provides a match score. This can cause the system to present errors given access to an incorrect person [ 59 ]. Environmental issues can also cause errors in the system. Temperature, humidity, and illumination conditions are some of the factors that can influence a biometric system to fail. In physiological recognition, there can be errors during the process, e.g., a student's voice can be altered due to sickness or a noisy environment during the recognition. Performance errors that can be mentioned include computing good quality images, the composition of the population, and the robustness of recognition algorithms [ 68 ].

3.4 Privacy issues

Biometric systems have the potential to collect a vast amount of information about people. Privacy is a concern that students and parents have regarding the use of biometric systems in schools. For example, finger and palm print biometric can track individual's dietary habits, which can be intrusive. In 2002, schools in the UK began to scan fingerprints of children without asking for parental consent. These are some of the events that make the academic community have fear mainly about the uses that the data collected will have [ 9 ]. However, there are biometric systems that do not necessarily store images of the characteristics being used. For example, in fingerprint recognition, a number is given to a student's fingertip when the first pass in the biometric device. The number is again generated for subsequent reads and compared with the numbers stored, identifying the student if there is a match [ 60 ]. In e-learning, where a person can learn at every moment, it is crucial to ensure that the personal data collected is safe and provides a personalized experience [ 15 , 16 ]. Therefore, biometric systems should incorporate privacy-protective solutions in their architecture. These should allow a system to work appropriately and give the users confidence and keep personal information safe [ 77 ].

3.5 Improvement of biometric technology

Biometric sensors should provide high-quality data that meet standards in mobile and challenging environments. Also, these should work by collecting data from faraway places, i.e., at a distance. Biometric systems should be increasingly accurate regardless of the environment in which they are doing [ 77 ]. Matching algorithms should also be improved, so they work in the presence of imperfect segmentation, noisy features, and signal variance. On the other hand, the technology should be robust enough to face attacks that include falsified biometric traits [ 51 ]. These systems are vulnerable due to their components, such as capture devices, communication channels, and databases. It is essential to improve these issues to work efficiently enough [ 62 ].

3.6 The new reality after CoViD19

Humanity before the Covid19 lockdown now seems to have passed a lifetime, and the psychological impact of what we are currently experiencing is likely to be enormous, even incalculable proportions. Facial recognition was promoted as one of the significant changes brought about by the coronavirus pandemic. To counter-balance identification problems due to partial face concealment, biometric Machine Learning and Artificial Intelligence algorithms are being pushed to the next evolutionary step [ 11 ].

CovidTech is a new technology and data movement due to the coronavirus pandemic. CovidTech constitutes the broadest and most profound period of agile technology and data development involving personal data processing.

There are two critical issues at CovidTech : epidemiology and exit strategies from the lockdown. Facial recognition is a crucial part of CovidTech . For example, buses in China have facial recognition systems that are combined with bus thermometers. Russia redesigned its facial recognition technology. Apps are being developed in Poland and India to allow people to upload selfies to contribute to the facial recognition system. Contact tracing is a proven epidemiological technique that has worked successfully in other health emergencies. These are only some examples of the CovidTech solutions that have potential in Higher Education Institutions , but there are certainly some contact-based Biometrics that no have room now. However, there are legal issues of use and scope that have not yet been resolved (Room, 2020).

3.7 Some societal and ethical issues

Technology improves the limits of our abilities but challenges our moral boundaries. Biometrics technology is part of our day; for example, through brain implants, social networks like Facebook, etc., using big data and emotion recognition, some systems learn to behave like humans [ 69 ].

Biometric systems allow the use of unique physical characteristics for verification or identification purposes. But, they invade our socio-cultural world, generating some social and ethical issues. Today, most public and political interests focus on privacy (especially personal data protection) and digital security. An analysis of the scientific literature on biometric systems revealed several recurring themes: privacy, security, autonomy, justice, human dignity, control of technology, and balance of powers. The various ethical and social problems manifest themselves in different ways. We will briefly comment on some examples to raise awareness of this issue (Royakkers et al., 2018).

Concerning privacy, biometric technology requires a minimum amount of information to determine if someone has the right to enter a building. On the other hand, because biometrics can identify sensitive data, controlling what happens to that information can be difficult, especially now that the technology has reached the stage of being applied in many more devices and situations. In the example of access to a building, the identity is not revealed; the verification can also be done by comparing someone's biometric characteristics with the information already stored about that person. Thus, biometrics is an excellent way to demonstrate legitimacy while maintaining privacy.

One application of biometrics is facial recognition and identification used in police investigations; this is regulated by law. This highly sensitive information must be stored securely, as biometric data can contain health (risk) and ethnicity. You don't want an insurance company to get hold of the information. An iris scan can determine diabetes or high blood pressure. Fingerprint irregularities can indicate leukemia or breast cancer.

The next generation of biometrics not only gives an idea of "who are you" but also answers the question “how do you feel” [ 45 ]. Emotion recognition technology provides information about people's mental state by examining people's automatic non-verbal comments without knowing it. For example, specific walking, grimacing, and other facial expressions can reveal something about a person and their behavior [ 17 ]. Unfortunately, a biometric system can misclassify and stigmatize someone as a terrorist, criminal, or unreliable individual. This can lead to revoking the presumption of innocence, this stigma likely stays with that person [ 61 ].

4 Biometric educational trends

Fingerprints, faces, and iris are well-known and consolidated biometric technologies; newer biometrics emerging for different applications include higher education. Although some technologies began to develop many years ago, they have not achieved the proper application level. A brief review is presented with some applications, Fig.  4 .

Biometrics educational trends

4.1 Vein/vascular pattern recognition

Vascular pattern recognition (vein pattern authentication) uses near-infrared light to reflect images of a person's blood vessels. The vascular pattern of humans is unique and does not change over the years. The potential for using this technology lies in obtaining trans body images and the possible applications of optical computed tomography scanning [ 56 ]. This technology uses subcutaneous blood vessels on the back of the hands or fingers due to their easy access since the near-infrared rays generated by a bank of LEDs can penetrate the skin on the back of the hand. Due to the difference in absorbance of blood vessels and other tissues, reflected near-infrared rays produce a vascular pattern extracted by various image processing techniques. Different feature-rich data (vessel branch points, vessel thickness, and branch angles) are extracted and stored as a template. Some disadvantages of this vein / vascular technology could be that (a) the quality of the captured image can be affected by body temperature, ambient temperature, humidity, uneven heat distribution, heat radiation, the proximity of the vein to the body surface, camera and focus calibration, and (b) this technology is invasive, which can be a significant limitation, especially in times of pandemic.

Compared to other biometric scanners, palm vein recognition technology has reduced false acceptance/rejection rates; improves the speed and security of endpoint contactless authentication [ 50 ], mainly due to the continuous development of the sensors used. A biometric finger vein scanner using infrared light and a CMOS image sensor in a specially designed system to reduce optical background noise in vein images are commercial solutions [ 8 ]. In addition, NASA developed a new subcutaneous structure imager that can locate veins in humans, Compared to the state of the art vein imaging; it has several characteristics [ 72 ].

4.2 Ear shape recognition

The use of ear imaging to identify people occurred more than 100 years ago. There are three methods of imaging the ear for identification: (a) taking a photograph, (b) taking marks, and (c) taking thermographic images of the ear. The essential parts of the ear for identification purposes correspond to the outer ear and the earlobe, but the structure and shape of the entire ear are also often necessary. Using the Image Ray Transform algorithm [ 74 ], a clear image of the outer ear can be systematically generated, this image is translated into a series of numbers for an identification process with a 99.6% success rate with several advantages such as (a) the ears are not affected by facial expressions (or the background), (b) the ears are consistent the age of the person, (c) ear imaging is non-invasive (can even be performed at long distance). However, some disadvantages include the transformation of the ears (with the use of jewelry or other accessories), the hair, the headphones, the lenses that cover the ears, the low light conditions, and the different angles of the ears images of ears.

Some commercial applications, such as the biometric system based on recognizing the ear canal's geometry to authenticate smartphone users with hearing aids, demonstrated efficiency of ~ 95% [ 7 ]. When a sound is played in someone's ear, the sound is propagated and reflected and absorbed through the ear canal, all of which produce a unique signature that can be recorded with the microphone [ 21 ]. Other commercial systems indicate that they can distinguish between people based on their acoustic characteristics with an accuracy greater than 99% [ 47 ].

4.3 Facial thermography

Facial thermograms are unique to people and could devise methods and systems for identifying them. Thermograms show the amount of infrared energy emitted, transmitted, and reflected by objects. This infrared energy is converted into temperature, which offers an image with this temperature distribution [ 53 ]. Thus, thermography can function as a facial recognition system in which an infrared camera is used to capture the images. In addition, facial thermogram technology is more accurate and robust in different lighting and environmental conditions than video images, using biosensor data to uniquely and automatically identify people.

Thermograms are a non-intrusive human identification technology in which facial recognition using multispectral imaging modalities, such as IR imaging sensors, has become an area of growing interest. Thermal imaging can avoid some common facial recognition challenges based on video images, such as (a) face and skin detection, location and segmentation are more comfortable to perform, (b) within-class variation. Furthermore, it is less since they are not affected by external distortions, (c) invariance to changes in lighting and facial expressions, (d) works in total darkness, and (e) detects costumes (even though the person has makeup).

Additionally, electronic thermography is used more frequently as a non-ionizing and non-invasive alternative for medicine diagnosis [ 36 ]. The vascular heat emissions present on a person's face can provide physiological indicators of the state of health or, where appropriate, the underlying disease.

4.4 Odor sensing

Identifying people through odors is not a new idea; Bloodhound dogs have been trained for years in this task, as body odor is a handy identifier. The primary odor contains stable components over time and does not change with food or the person's environment, making it a successful biometric. A person's secondary odor includes characteristics that are present due to dietary or environmental factors. Finally, the tertiary odor contains elements present due to external factors such as soaps and perfumes.

Real sensors have not yet reached the precision of a dog's smell. Thanks to nano and biotechnology, there are still fundamental advances (thanks to nano and biotechnology) with high sensitivity to detect volatile elements in people's body odor [ 37 ]. Even body odor can vary due to illness, food, mood swings, etc. These recognizable patterns in the body odor of each person have an identification error of 15%. This biometric technology is not intrusive, it works even with deodorants and perfumes since it is impossible to reproduce human's smell. Electronic nose research groups have developed some prototypes of noses to smell different odors and aroma types [ 63 ], some have been successfully tested to search for people in risky situations such as earthquakes or other natural disasters. People can also be identified through their unique breath prints; This method is fast and non-invasive [ 43 ].

4.5 Gait recognition

Gait recognition is a behavioral biometric technology that identifies people based on their unique gait patterns. Minor variations in gait style can be used as a biometric identifier to identify individuals. Gait recognition correlates Spatio-temporal parameters such as stride length, stride width, gait speed, and cycle time with kinematic parameters such as joint rotation and the joint's average angles, hip, knee and ankle, and the angles of the thigh, trunk, and foot; The length of the stride and the height of the person are also considered. Some research projects show 95% efficiency in identifying people [ 76 ].

There are important commercial applications that developed a gait recognition technique based on a smartphone's information using a dynamic temporal distortion classification algorithm and an artificial neural network as an evaluation system to authenticate people [ 46 ]. Significant research investments have been generated to bring this image-based gait analysis to the market [ 44 ]. However, it is not the only case, several gait recognition systems have been developed, considering it an excellent biometric [ 6 ].

4.6 Heartbeat authentication

Heartbeat can not only detect humans, but it can also identify them. This is because the heartbeat follows an irregular pattern that never repeats and is unique to each person [ 3 ]. With an electrocardiograph (ECG), the unique mathematical features underlying this pattern can be extracted, and a secret key generated the system can use that to encrypt the information. Small changes in the initial conditions lead to very different results [ 12 ]. The shape of the heartbeat pattern is affected by several factors (size, shape, and heart position in the body).

The advantages of biometric information are: (a) it is difficult to clone a biometric marker or an ECG, (b) an ECG cannot be recorded without the permission of a person, and (c) it is non-invasive [ 66 ]. On the other hand, some disadvantages are (1) difficulty in capturing heartbeat patterns, (2) transformation with age or exercise response.

NASA has released a new patent for a heartbeat-based biometric system for commercial licenses to enable solution developers and integrators [ 72 ]. Besides, Huang et al. [ 31 ] strategically reuse ECG signals to improve data encryption, security level, and privacy.

4.7 Brain waves/activity

Using brain waves as an identifier is relatively new biometrics; it could serve as a security system and verify a person's identity with less than 1% [ 1 ]. However, the noise associated with all the brain signal measurements has made data analysis difficult. Focusing on the waves in the brain area related to reading and word recognition produces a clearer signal that can be measured faster [ 30 ].

These brain signals are generated when people access their semantic memories, which record specific word's meanings. This method could become a more personal and secure authentication alternative [ 39 ]. Additionally, scanned images of brain activity using functional magnetic resonance imaging can act as a signature pattern to accurately identify people. Some techniques measure neural activity through blood flow in the brain while people are awake and mentally active [ 65 ]. One advantage is that it is almost impossible to clone. The downsides are that the technology is intrusive and not easy to use because it is specialized equipment.

4.8 Human body bioacoustics

A modulated micro-vibration that travels through our body could capture a person's unique spectral trait [ 49 ]. Sound waves passing through the human body can identify a person with 97% accuracy, analyze properties such as the individual's skin, joint tension, and bone density [ 57 ]. The spectral patterns of bioacoustics do not undergo significant changes over time. Researchers used a transducer to generate vibrations and sound waves that pass through a person's body. The finger is a part of the body that has been used to test this technique. Once sound passes through skin, bone, and other tissues, a sensor picks up the unique bioacoustic signature.

The biometric system can be very successful in identifying people; for example, different individual fingers can be identified. This means that a person must use the same finger used for authentication [ 27 ]. Auditory acoustics can also be used for authentication. The ear has unique characteristics that change from person to person [ 2 ].

4.9 Perceptual user interface

It is possible to use gestures to perform human–machine interaction with gesture recognition, a mathematical interpretation of human movements through a computational device. Currently, computing devices and smartphones have gesture recognition systems, games, and virtual reality systems, more interactive and immersive (i.e., Microsoft's Kinect). These systems can also be used to authenticate a user [ 63 ].

Cheiloscopy is a forensic investigation technique that deals with the identification of humans from lip traces. Lip prints are unique and also permanent for every human being. Lip prints between family members (and between twins) generally revealed different patterns with few similar grooves suggesting inheritance in lip prints [ 70 ]. Cheiloscopy has mainly served forensics as an aid for personal identification; This technology has an enormous development [ 63 ].

4.10 New platforms

Biometric technology (as well as traditional systems) has been used mainly when the authentication or verification of the person occurs. However, new platforms are growing due to new demands and developing services. Mobile devices are growing, evolving, incorporating the latest technology platforms, and, with this, biometric systems are transforming and expanding. For example, Samsung's Galaxy Tab has incorporated iris recognition technology to access the device. In addition, some portable devices, such as bracelets, measure the user's heartbeat to authenticate purchases made with Mastercard credit cards on the site [ 22 ].

5 Conclusions

Biometric technology is relatively new and has changed the way identification and authentication processes are performed. It comprises innovative advances that measure the physical and behavioral characteristics of people. Biometrics is now used in various fields, including government, banking, healthcare, and education. The latter has reaped the benefits of biometric options to good effect. In addition, biometrics can be used to perform non-academic and academic activities at universities. Non-academic activities include the use of fingerprints to enter the campus or university dormitories, pay for a cafeteria or library service. Educational activities include monitoring student behavior or emotions to change teaching strategy on time, class attendance, taking electronic tests, analyzing motivations and progress in a course, and conducting learning analysis. However, some challenges need to be solved effectively, including user's reluctance to use technology. People may be afraid of how their personal information will be used.

As this is a new field of study, it could be challenging to provide students with the knowledge to develop such systems. In addition, privacy concerns are the primary concern, as devices can collect all kinds of personal information. Likewise, there is a need to improve biometric technology to make it more robust and develop more innovative solutions that solve society's main problems or add value to the services offered, in this case, in the educational field.

An important aspect to consider is the choice of the appropriate biometric solution for educational institution's needs. Some factors to consider are (Fry and Dunphy, nd): (a) choose a biometric identification system that can be used in multiple areas, that is, students are identified only once and have access to various services at the university (access to campus, cafeteria, classroom) attendance, library, etc.). (b) the biometric system must be scalable, that is, it must work well with a handful of students and increase their number, (c) the system must be compatible with the software that is already installed on campus, (d) before starting the system, make sure it works well; all traits of potential users need to be scanned, (e) consider accuracy and performance, (f) compare different providers, so that selection is based on an informed process, (g) communicate how the system works to all who will use it.

The biometric market is expected to reach $ 94 billion in 2025, at a compound annual growth rate of 36%, over the next 5 years. New features are being developed, such as vascular pattern recognition, ear shape recognition, facial thermography, odor detection, gait recognition, authentication of heartbeat, brain waves, and bioacoustics of the human body. The development of typical artificial intelligence algorithms combined with the technological growth of the Internet of Things (and its impact on Big Data) will be great catalysts for biometric systems to develop widely and impact many areas within them, including education. However, the biggest challenge for this technology to overcome is security and privacy concerns; they must be addressed to harness this technology's full potential.

Finally, biometrics is a technology that can enhance students’ abilities in the context of Industry 4.0. Various interactive technologies fulfill this task. Such technologies are (a) 3D printing which enhances students understanding; (b) Augment reality which is an interactive technology that captures students attention; (c) Virtual Reality enhances students ability to connect theory and practice; (d) Cloud computing can be used to have access to the learning material from anywhere; (e) Holograms offer interactive spaces to learn [ 26 ]; (f) Artificial Intelligence permits students to have a better understanding of a given subject; (g) Robots capture students attention [ 67 ]; (h) Internet of Things allow students to be active learners [ 42 ].

Anthony, S.: No TitleBerkeley researchers replace passwords with pass thoughts by reading your mind. Retrieved April 15, 2020, from www.extremetech.com/computing/152827-berkeley-researchers-authenticate-your-identity-with-just-your-brainwaves-replace-passwords-with-passthoughts (2013)

Arakawa, T.: Ear acoustic authentication technology: using sound to identify the distinctive shape of the ear canal. Retrieved July 10, 2020, from www.nec.com/en/global/techrep/journal/g18/n02/180219.html (2018)

Aron, J.: Your heartbeat could keep your data safe. Retrieved April 15, 2020, from www.newscientist.com/article/mg21328516.500-your-heartbeat-could-keep-your-%0Adata-safe/%0A (2012)

Baird, S.: Biometrics “Security Technology”: it is important for students to understand that technology can be used as part of a solution to a problem. Technol. Teach. 61 (5), 18–23 (2002)

Google Scholar  

Bowyer, K.W., Hollingsworth, K., Flynn, P.J.: Image understanding for Iris biometrics: a survey. Comput. Vis. Image Underst. 110 , 281–307 (2008)

Article   Google Scholar  

Burt, C.: Researchers unveil new AI gait recognition system. Retrieved from www.biometricupdate.com/201805/researchers-unveil-new-ai-gait-recognition-system (2018)

Burt, C.: Biometric ear canal geometry recognition developed by University of Buffalo Researchers. Retrieved April 15, 2020, from www.biometricupdate.com/201909/biometric-ear-canal-geometry-recognition-developed-by-university-of-buffalo-researchers (2019a)

Burt, C.: New technologies for finger vein, voice, and facial biometrics unveiled. Retrieved April 13, 2020, from www.biometricupdate.com/201908/new-technologies-for-finger-vein-voice-and-facial-biometrics-unveiled (2019b)

Buttle, C.: The Problem of Biometrics in Education. Biometric Technol. Today 2013 (6), 5–7. Retrieved from www.sciencedirect.com/science/article/abs/pii/S0969476513701115 (2013)

Caldwell, T.: Biometric tech has mixed fortunes in education. Biometric Technol. Today 3 , 12 (2013)

Carlaw, S.: Impact on biometrics of Covid-19. Biometric Technol. Today 8–9 (2020)

Chen, C., Lin, C., Chiang, C., Lin, S.: Personalized information encryption using ECG signals with chaotic functions. Inf. Sci. 193 , 125–140 (2012)

Dafoulas, G.A., Maia, C.C., Clarke, J.S., Ali, A., Augusto, J.: Investigating the role of biometrics in education—the use of sensor data in collaborative learning. In: Int Conf e-Learning, pp. 115–123 (2018)

Dell Technologies Inc.: Gen Z is here. Are you ready? (2020)

Dinu, V.E., Papuc, D., Gheorghiu, A., Dascalu, M.: Biometric data in learning analytics: a survey on existing applications. In: The 13th Int Scientific Conf eLearning and Software for Education, pp. 465–473 (2017)

Dinu, V., Papuc, D., Gheorghiu, A., Dascalu, M. L., Moldoveanu, A., & Moldoveanu, F.: Biometric data in learning analytics: a survey on existingcapplications. In: 13th Int Scientific Conf eLearning and Software for Educatioc Bucharest, pp. 465–472 (2017)

Dwoskin, E., Rusli, E.M.: The technology that unmasks your hidden emotions. Wall Street Journal (Jan 28). Accessed Oct 20, 2020, from wsj.com/articles/startups-see-yourface-unmask-your-emotions-1422472398#:bLk8dH_DkLSJvA (2015)

Education Publishing Worldwide Ltd: Biometrics an option for schools wanting a secure future. Educ. J 280 , 27 (2016)

Elliott, S.J., Peters, J.L., Rishel, T.J.: An introduction to biometrics technology: its place in technology education. J. Ind. Teach. Educ. 41 (4), 1–8 (2004)

Ferguson, R.: Learning analytics: drivers, developments and challenges. Int. J. Technol. Enhanced Learn. 4 (5/6), 304–317 (2012)

Gao, Y., Wang, W., Phoha, V. ., Sun, W., Jin, Z.: EarEcho. In: ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, p. 1 (2019)

German, R., Barber, K.S.: Current biometric adoption and trends. The University of Texas at Austin. Retrieved from identity.utexas.edu/assets/uploads/publications/Current-Biometric-Adoption-and-Trends.pdf (2017)

Gold, S.: Biometrics in education: integrating with the real world. Biometric Technol. Today 2010 (4): 7–8. Retrieved from www.researchgate.net/publication/238294356_Biometrics_in_education_integrating_with_the_real_world (2010)

Grawemeyer, B., Mavrikis, M., Holmes, W., Gutierrez-Santos, Wiedmann, M., Rummel, N.: Affecting O-task behaviour: how affect-aware feedback can improve student learning. In: LAK, pp. 104–113 (2016)

Gray, S.L.: Biometrics in schools: the role of authentic and inauthentic social transactions. UCL Institute of Education. Retrieved from discovery.ucl.ac.uk/id/eprint/1545213/3/Leaton Gray_Biometrics BSA conference.pdf (2017)

Halili, S.H.: Technological advancements in Education 4.0. Online J. Distance Educ. E-Learn. 7 (1), 63–69 (2019)

Hampson, M.: The bioacoustic signatures of our bodies can reveal our identities. Retrieved July 10, 2020, from spectrum.ieee.org/the-human-os/telecom/security/the-bioacoustic-signatures-of-our-bodies-can-reveal-our-identities (2019)

Hidrogo, I., Zambrano, D., Hernandez-de-Menendez, M., & Morales-Menendez, R.: Mostla for engineering education: part 1 initial results. Int. J. Interact. Des. Manuf. https://doi.org/10.1007/s12008-020-00730-4 (2020)

Hoffman, F.: The effectiveness of biometrics in student education. Retrieved from www.m2sys.com/blog/guest-blog-posts/the-effectiveness-of-biometrics-in-student-education/ (2020)

Hond, B.: Your brain’s unique response to words can reveal your identity. Retrieved April 15, 2020, from www.newscientist.com/article/dn27555-your-brains-unique-response-to-words-can-reveal-your-identity/ (2015)

Huang, P., Li, B., Guo, L., Jin, Z., & Chen, Y.: A robust and reusable ECG-based authentication and data encryption scheme for eHealth systems. In: 2016 IEEE Global Comm Conf, pp. 1–6 (2016)

Jager, T.D.: Application of biometric fingerprinting to encourage the active involvement of student teachers in lectures on differentiated instruction. South Afr. J. Educ. 39 (2), 1–11 (2019)

Jain, A., Ross, A., Prabhakar, S.: An introduction to biometric recognition. IEEE Trans. Circuits Syst. Video Technol. 14 , 4–20 (2004)

Jraidi, C.M., Frasson, C.: A hierarchical probabilistic framework for recognizing learners’ interaction experience trends and emotions. In: Advances in Human–Computer Interaction (2014)

Kim, K.H., Bang, S.W., Kim, S.R.: Emotion recognition system using short-term monitoring of physiological signals. Med. Biol. Eng. Comput. 42 (3), 419–427 (2004)

King, R.: Explainer: Facial Thermography. Retrieved April 15, 2020, from www.biometricupdate.com/201308/explainer-facial-thermography (2013)

Korotkaya, Z.: Biometric Person Authentication: Odor. Department of Information Technology, Laboratory of Applied Mathematics, Lappeenranta University of Technology (2003)

Krucoff, M.O., Rahimpour, S., Slutzky, M.W., Reggie, E.V., Turner, D.A.: Enhancing nervous system recovery through neurobiologics, neural interface training, and neurorehabilitation, neuroprosthetics. Front. Neurosci. 10 , 584 (2016)

Lee, J.: Researcher says brainwaves could be a way for security systems to verify identity. Retrieved April 15, 2020, from http://www.biometricupdate.com/201505/researcher-says-brainwaves-could-be-a-way-for-security-systems-to-verify-identity (2015)

Levy, Y., Ramim, M.: A Theoretical Approach for Biometrics Authentication of E-exams. Nova Southeastern University, USA (2007)

Lynn, A.: Cyber Attacks History In Higher Education. Retrieved April 15, 2020, from www.informationsecuritybuzz.com/articles/cyber-attacks-history-in-higher-education/ (2018)

Maksimovic, M.: Facta universitatis, series: teaching, learning and teacher education. Facta Univ. Ser. Teach. Learn. Teach. Educ. 1 (2), 137–150 (2018)

Martinez-Lozano, P., Kohler, M., Zenobi, R.: Human breath analysis may support the existence of individual metabolic phenotypes. PLoS ONE 8 (4), e59909 (2018)

Mayhew, S.: Watrix Raises $14M, Releases Gait Recognition Solution. Retrieved April 15, 2020, from www.biometricupdate.com/201810/watrix-raises-14m-releases-gait-recognition-solution (2018)

Mordini, E., Tzovaras, D., & Ashton, H.: Introduction. In: Mordini, E., Tzovaras, D. (Eds.), Second generation biometrics: the ethical, legal and social context. The International Library of Ethics, Law and Technology, vol. 11, pp. 1–19. Dordrecht: Springer (2012)

Mufandaidza, M., Ramotsoela, T., Hancke, G.: Continuous user authentication in smartphones using gait analysis. In: 44th Annual Conf of the IEEE Industrial Electronics Society (2018)

NEC: NEC develops biometrics technology that uses sound to distinguish individually unique ear cavity shape. Retrieved April 15, 2020, from www.nec.com/en/press/201603/global_20160307_01.html (2016)

Nita, S., Mihailescu, M.: Proposing a secure e-learning system based on biometric authentication and homomorphic encryption. In: 14th Int Scientific Conf eLearning and Software for Education Bucharest, pp. 47–52 (2018)

Pascu, L.: Human body bioacoustics deliver 97 percent biometric accuracy, researchers say. Retrieved April 15, 2020, from www.biometricupdate.com/201911/human-body-bioacoustics-deliver-97-percent-biometric-accuracy-researchers-say (2019)

Pascu, L.: IGEL Integrates Fujitsu palm vein recognition to boost endpoint security. Retrieved April 13, 2020, from www.biometricupdate.com/202004/igel-integrates-fujitsu-palm-vein-recognition-to-boost-endpoint-security (2020)

Pato, J.N., Millett, L.I.: Biometric Recognition. Challenges and Opportunities. The National Academies Press. Retrieved from dataprivacylab.org/TIP/2011sept/Biometric.pdf (2010)

Pearson: Meeting the Expectations of Gen Z in Higher Education. Retrieved July 6, 2020, from www.pearson.com/us/content/dam/one-dot-com/one-dot-com/us/en/files/PSONA5646-8150_TIDL_GenZ_Infographic_Print_FINAL.pdf (2018)

Prokoski, F., Riedel, R.: Infrared identification of faces and body parts. In: Biometrics: Personal Identification in Networked Society. Springer, USA (1996)

Russell, J.A.: A circumplex model of affect. J. Person. Soc. Psychol. 39 (6), 1161–1178 (1980)

Salmeron-Majadas, S., Arevalillo-Herráez, M., Santos, O. C., Saneiro, M., Cabestrero, R., Quirós, P., Boticario, J.G.: Filtering of spontaneous and low intensity emotions in educational contexts. In: AIED, pp. 429–438 (2015)

Shimizu, K.: Optical trans-body imaging—feasibility of optical CT and functional imaging of living body. Med. Philos. 11 , 620–629 (1992)

Sim, J., Noh, H., Goo, W., Kim, N., Chae, S., Ahn, C.: Identity recognition based on bioacoustics of human body. IEEE Trans Cybern 1–12 (2019)

Society of Learning Analytics: About. Retrieved July 6, 2020, from https://www.solaresearch.org/ (2020)

Šošević, U., Milenković, I., Milovanović, M., Minović, M.: Support platform for learning about multimodal biometrics. J. Univers. Comput. Sci. 19 (11), 1684–1700 (2013)

Spalding, S.: Biometrics in tertiary education. Biometric Technol. Today 2011 (5), 8–9 (2011)

Sutrop, M., Laas-Mikko, K.: From identity verification to behavior prediction: ethical implications of second generation biometrics. Rev. Policy Res. 29 (1), 21–36 (2012)

Syed, S., Ali, B., Wan, W.: Technical issues and challenges of biometric applications as access control tools of information security. Int. J. Innov. Comput. Inf. Control 8 (11), 7983–7999 (2012)

Thakkar, D.: Five Unconventional Biometrics that Surprisingly Exist! Retrieved April 15, 2020, from https://www.bayometric.com/5-unconventional-biometrics/ (2018)

Thales: Biometrics: authentication and identification (definition, trends, use cases, laws and latest news)—2020 review. Retrieved July 7, 2020, from https://www.thalesgroup.com/en/markets/digital-identity-and-security/government/inspired/biometrics (2020)

The Conversation: Brain Activity is as Unique—and Identifying—as a Fingerprint. Retrieved April 15, 2020, from https://theconversation.com/brain-activity-is-as-unique-and-identifying-as-a-%0Afingerprint-48723.%0A (2015)

The Economist: A Heart to my Key. Retrieved April 15, 2020, from http://www.economist.com/blogs/babbage/2013/05/biometrics (2013)

Timms, M.J.: Letting Artificial Intelligence in Education out of the Box: Educational Cobots and Smart Classrooms. Int. J. Artif. Intell. Educ. 26 (2), 701–712 (2016)

Unar, J.A., Chaw, W., Abbasi, A.: A review of biometric technology along with trends and prospects. Pattern Recogn. 47 (8), 2673–2688 (2014)

Van Est, R.: Intimate Technology: The Battle For Our Body And Behaviour. Rathenau Instituut, The Hague (2014)

Venkatesh, R., David, M.: Cheiloscopy: an aid for personal identification. J Forensic Dent Sci. 3 (2), 67–70 (2011)

Verizon: 2019 Data Breach Investigations Report. Retrieved April 15, 2020, from https://enterprise.verizon.com/resources/reports/dbir/ (2019)

Vincent, B.: The Space Agency Announced Several Health and Biotech Technology Transfer Opportunities. Retrieved April 13, 2020, from https://www.nextgov.com/emerging-tech/2020/02/nasa-tech-could-replace-passwords-your-heartbeat/163292/ (2020)

Wampfler, R., Klingler, S., Solenthaler, B., Schinazi, V.R., Gross, M.: Affective state prediction in a mobile setting using wearable biometric sensors and stylus. In: Proc of The 12th Int Conf on Educational Data Mining, pp. 198–207 (2019)

Yale Scientific: Ears: the new fingerprints? Retrieved April 13, 2020, from www.yalescientific.org/2011/05/ears-the-new-fingerprints/ (2011)

Yampolskiy, R.V., Ave, M.: Behavioural biometrics : a survey and classification. Int. J. Biometrics 1 (1), 1–33 (2008)

Yang, K., Dou, Y., Lv, S., Zhang, F., Lv, Q.: Relative distance features for gait recognition with kinect. J. Vis. Commun. Image Represent. 39 , 209–217 (2016)

Yang, L., Winters, K., Kizza, J.M.: Biometrics education with hands-on labs. In: Proc of the 46th Annual Southeast Regional Conference, pp. 18–23 (2008)

Download references

Acknowledgements

The authors would like to acknowledge the financial and technical support of Writing Lab, TecLabs, Tecnológico de Monterrey , México, in the production of this work.

Author information

Authors and affiliations.

Tecnológico de Monterrey, Av. E Garza Sada 2501, 64849, Monterrey, NL, Mexico

Marcela Hernandez-de-Menendez & Ruben Morales-Menendez

General Motors, Global Research and Development, Warren, MI, USA

Carlos A. Escobar & Jorge Arinez

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Ruben Morales-Menendez .

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Hernandez-de-Menendez, M., Morales-Menendez, R., Escobar, C.A. et al. Biometric applications in education. Int J Interact Des Manuf 15 , 365–380 (2021). https://doi.org/10.1007/s12008-021-00760-6

Download citation

Received : 26 November 2020

Accepted : 22 July 2021

Published : 28 July 2021

Issue Date : September 2021

DOI : https://doi.org/10.1007/s12008-021-00760-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Educational innovation
• Higher education
• State of the art
• Find a journal
• Publish with us
• Track your research

Help | Advanced Search

Computer Science > Cryptography and Security

Title: advancements in biometric technology with artificial intelligence.

Abstract: Authentication plays a significant part in dealing with security in public and private sectors such as healthcare systems, banking system, transportation system and law and security. Biometric technology has grown quickly recently, especially in the areas of artificial intelligence and identity. Formerly, authentication process has depended on security measures like passcodes, identity fobs, and fingerprints. On the other hand, as just a consequence of these precautions, theft has increased in frequency. In response, biometric security was created, in which the identification of a person is based on features derived from the physiological and behavioral traits of a human body using biometric system. Biometric technology gadgets are available to the public as they are embedded on computer systems, electronic devices, mobile phones, and other consumer electronics. As the fraudulent is increasing demand and use of biometric electronic devices has increased. As a consequence, it may be possible to confirm a person's distinct identification. The goal of this study is to examine developments in biometric systems in the disciplines of medicine and engineering. The study will present the perspectives and different points of view of the secondary data, highlighting the need for more in-depth understanding and application of biometric technology to promote its development in the digital era. The study's findings may inspire people and businesses to more effectively incorporate biometric technologies in order to reduce the risks to data and identity security.

Submission history

Access paper:.

• Other Formats

References & Citations

• Google Scholar
• Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

• Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• PeerJ Comput Sci

Hand-based multibiometric systems: state-of-the-art and future challenges

1 National University of Sciences and Technology, Islamabad, Pakistan

Farrukh Aslam Khan

2 Center of Excellence in Information Assurance (CoEIA), King Saud University, Riyadh, Saudi Arabia

Muhammad Khurram Khan

Haider abbas, waseem iqbal, farhan riaz, associated data.

The following information was supplied regarding data availability:

There is no raw data/code in this paper as it is a survey paper.

The traditional methods used for the identification of individuals such as personal identification numbers (PINs), identification tags, etc., are vulnerable as they are easily compromised by the hackers. In this paper, we aim to focus on the existing multibiometric systems that use hand based modalities for the identification of individuals. We cover the existing multibiometric systems in the context of various feature extraction schemes, along with an analysis of their performance using one of the performance measures used for biometric systems. Later, we cover the literature on template protection including various cancelable biometrics and biometric cryptosystems and provide a brief comment about the methods used for multibiometric template protection. Finally, we discuss various open issues and challenges faced by researchers and propose some future directions that can enhance the security of multibiometric templates.

Introduction

Biometric authentication is used more than ever for authentication of individuals in a wide range of security applications. The reliance of systems on physiological attributes of the users has lately offered more simplicity and reliability at the same time. This has helped in avoiding many problems associated with the systems where passwords/credentials are being used, which can potentially incur some problems such as forgotten passwords, transferred or stolen credentials. The use of biometrics has led to mitigate these problems significantly given that the individual with specific biometric traits is required to validate access to the systems to avoid the above mentioned problems. Moreover, most of the existing systems are typically connected to networks, at the very least, a local area network connecting a local network with a couple of systems and more often, a wide area network eventually connecting to the World Wide Web. Given this, a protection mechanism is required to be in place to ensure that an unauthorized access to the system is prevented and the templates are properly protected.

There is a wide use of authentication systems in Internet services and mobile devices for the protection of the user content. Various tools and techniques for the management of information security have been developed. However, systems based on biometrics have made significant progress to support some aspects of information security over the period of time. An in-depth and comprehensive study on biometric authentication has been conducted in recent years by various researchers ( Jain, Ross & Prabhakar, 2004 ; Jain, Nandakumar & Ross, 2016 ). With the passage of time, biometric authentication of the users is gaining more and more popularity since the systems based on biometrics are not easily compromised. This is because, the systems can be breached only if the individuals who are trying to access the systems are in possession of those physiological parameters, which are possessed by the actual users. This has led to the addition of security for the protection of the systems, and reduced their vulnerability.

It goes without saying that the field of biometrics is very rich and up to the minute. There are a number of surveys that exist on biometric systems. However, some surveys focus mainly on a particular modality or environment; some of the recent contributions include Connor & Ross (2018) , in which the authors focused on a biometric recognition system based on gait. They have reviewed several gait recognition modalities and their features. Kumari & Seeja (2019) provides an in depth survey about periocular biometrics, using various existing feature extraction methods and matching schemes. The paper also emphasizes the importance of periocular biometrics in a wide range of applications. Dargan & Kumar (2020) have done a very comprehensive and in depth survey on various unimodal and multimodal biometric recognition systems discussing feature extraction methods, various classifiers and datasets. Their main aim is to make the researcher aware of multiple dimensions to look for in a biometric system in order to enhance its security. Sundararajan & Woodard (2018) have performed a survey on the use of deep learning in the domain of biometric authentication using various modalities. However, their conclusion is that most of the deep learning approaches have been explored mainly on face biometrics and speaker recognition. Dinca & Hancke (2017) emphasized the importance of multibiometric systems in their work for meeting the emerging security demand in the field of authentication. Their work is mainly focused on covering two important aspects in biometric systems: fusion methods and security. A thorough review of secure and privacy preserving authentication is presented in Rui & Yan (2018) . The authors have mainly tackled the problem of liveness detection and privacy protection in biometric systems. Given an ever increasing work done in the area of wearable technology and IoTs, the wearable biometrics is another upcoming area that requires a significant attention of the researchers. In this context, Sundararajan, Sarwat & Pons (2019) have performed an interesting review in which the authors compare the key characteristics of different modalities, and highlight critical attacks being carried out in traditional and wearable biometric systems. However, the scope of the manuscript is quite broad with a review covering most of the biometric modalities including behavioral and physiological traits. Moreover, there is a lack of presentation of the quantitative analysis in several manuscripts creating a gap for a more focused and thorough review on hand based multibiometric systems.

Rationale for the survey

There is a wide range of biometric traits that can be used for authentication purposes including face, hands, iris, retina, etc. All these traits offer their own advantages and drawbacks and most of them are thoroughly covered in the literature. In this paper, we aim to focus on the existing hand based multibiometric systems explored over the past five years. The use of hand modalities offers several advantages over others: they are highly accurate for recognition, generally make use of inexpensive technology, are fast for matching and require templates of very small sizes, resulting in a small memory footprint and are less sensitive to imaging conditions. Moreover, the hand-based modalities are more robust since they are not affected by emotions and other behavioral characteristics of the individuals such as tiredness, stress, etc. Given this, it is clear that with respect to some specific aspects, the use of hand based modalities is superior as compared to others for biometric authentication.

There is a large volume of literature discussing about the use of hand based modalities for authentication but the work becomes limited as it is directed towards multimodal systems. This domain was constrained until recently due to the issues related to the power consumption, size, and cost, etc., of the hardware required for executing the biometric systems. However, over the recent years, the revolution in the hardware design industry has led to miniaturized devices having tremendous capabilities, enabling the developers to execute multiple systems on very light, low-powered, and small devices with significantly lesser cost. Consequently, it has been possible to implement multibiometric systems very efficiently, triggering a lot of research in this direction. Moreover, to the best of our knowledge, a thorough survey of hand based multibiometric systems, their effective usage in biometric authentication and the main challenges faced, is currently not available in the literature.

With regard to the security of the biometric system, being multibiometric in nature adds itself another layer of security even though there are multiple points of attack on an authentication system. We aim to tackle the literature available regarding security of multibiometric templates. The rationale for focusing on security of multibiometric template is that they lead to a 3-dimensional vulnerability to a biometric system in contrast to their counterparts ( Jain, Nandakumar & Nagar, 2008 ): 1. Template can be replaced by an imposter to gain unauthorized access, 2. A spoof can be created from the template to aide in unauthorized access, and 3. The stolen template can be replayed to the matcher to gain access. Therefore, it is vital to protect the templates from an adversary; unlike PINs and passwords, a biometric template if compromised cannot be revoked and reissued, so considering the criticality in this context, we aim to deal with the research contributions that are devised to protect the integrity of the saved templates.

This paper presents a systematic review on the use of hand based multibiometric systems and an analysis of their efficacy in performing authentication. In this context, the main contributions are as follows:

• Discussing the main advantages and motivations behind the usage of hand based multibiometric systems.
• Presentation of a taxonomy to categorize the literature with respect to the two parameters: authentication and template protection.
• Presentation of a summary of literature on the above-mentioned parameters, with critical/brief comments.
• A discussion about open issues and the direction of future work on hand based multibiometric systems.

In summary, this paper covers two major directions of work on hand based multibiometric systems: 1. The work on various schemes to perform feature extraction and authentication of individuals using multibiometric systems based on hand modalities. The main objective of such studies is to ensure that the performance metrics of the systems are very good and they can be effectively used, and 2. Once the templates have been acquired from the users (the users are enrolled), how these templates can be effectively archived such that they are not susceptible to attacks.

Paper organization

The rest of the paper is organized as follows: In “Survey Methodology”, we present the methodology followed in this survey. We discuss an overview of a biometric system, multibiometric system and its types in “Overview of a Biometric System”, followed by the discussion on fusion methods used in a multibiometric system in “Fusion Methods”. Later, we talk about different hand-based modalities that are used for biometric authentication in “Hand Based Modalities”. Then, we discuss the feature extraction methods in existing hand multibiometric systems in “Feature Extraction Techniques for Hand Multibiometric Systems” and methods used to perform multibiometric template security in “Multibiometric Template Security”. Finally, we discuss about various open issues and challenges in the topics covered in “Open Challenges and Future Directions” and conclude the paper in “Conclusions”.

Survey methodology

With an ever growing demand of designing authentication systems and the linkages of such systems with critical databases owned by the governments, corporates and various entities, there is an increasing demand on making these systems scalable, user interactive, safe, and secure. In this context, the biometric technology has significantly grown over the last decade. Subsequently, several works have been conducted listing the major contributions and breakthroughs in the area. However, to the best of our knowledge, there is a shortage of detailed surveys on the use of hand-based modalities for multibiometric systems and an analysis of security aspects with respect to template protection. The recent contribution on this topic was done by Bahmed & Mammar (2019) . However, the survey is limited and lacks a thorough analysis and discussion on future directions and does not cover the security aspect of the biometric systems. Moreover, there is a need to define a clear taxonomy that helps in defining the future research directions in the subject area.

In this survey, the approach followed to collect the manuscripts is shown in Fig. 1 . The scholarly databases used for searching articles were IEEE explore, ACM Digital Library, Springer, Science Direct and Google Scholar. Most of the identified research papers are published in reputed forums. We have mainly focused on papers published from 2010 till date (May 2021). The search terms used for collecting the manuscripts were: “Multibiometric systems” with permutations of different hand modalities. According to the search terms, more than 600 papers were initially identified which were screened to fit the scope of the topic based on their title, abstracts, body, and conclusions. For our work on authentication using hand based multibiometric systems, we have identified 35 manuscripts, whereas for multibiometric template protection, we have picked out 22 manuscripts. We have followed a semi-systematic methodology for this survey ( Snyder, 2019 ), narrowing down the literature in multiple phases. Breadth search is first adopted in which the literature on all hand multibiometric systems is analyzed. It was concluded that the optimization parameters on multibiometric systems revolved around two important performance primitives: authentication and security. These are complementary parameters as optimizing security may compromise on the authentication results obtained for the systems, although, a vice versa may not be necessarily true. Therefore, this led us to the Phase II of our research in which we conduct a depth search to shortlist the literature based on the design of authentication of multibiometric systems and template security. Phase II is formalized as a taxonomy in Phase III, which structures the literature for a better understanding and comprehension of the underlying problem. The inference drawn from an evaluation of the literature has led to a discussion about some challenges leading to subsequent future directions of work in this domain (Phase IV).

Overview of a biometric system

In a biometric system, an identifier is linked to its intrinsic human characteristics. These characteristics are physiological and behavioral in nature, which can be used to identify a person digitally ( Meng et al., 2014 ; Rui & Yan, 2018 ). Biometric security helps in authentication, which takes place by identifying human characteristics. The specific human characteristics mentioned above are defined as follows:

• Physiological: Physiological biometrics are based on physical characteristics of an individual. They vary from individual to individual and are assumed to be relatively unchanging such as fingerprints, face, iris/retina etc.
• Behavioral: Behavioral biometrics are based on behavioral characteristics of an individual. The examples include voice, gait, signature, etc.

There are four important modules in a traditional biometric system ( Fig. 2 ). The sensor module is responsible to acquire data from the users, whereas the feature extraction module processes the sensor data to find a description that is feasible for matching of templates that are residing in the database. The feature extraction module is followed by the matching module that generates the matching scores, which are finally used to perform the decision making regarding the grant of permissions to a specific user in the decision module.

The several factors that are considered significant while performing the selection of a specific biometric identifier include permanence, universality, measurability, circumvention, performance, etc ( Bhattacharyya et al., 2009 ). Another important factor is the suitability of the application. Nevertheless, the choice of a single biometric identifier which meets all the requirements of every possible application is not possible since there are tradeoffs between different performance metrics. Keeping this in view, there is a possibility to optimize a number of measures by using a combination of various biometric identifiers. Therefore, we can logically characterise a biometric system into two distinct categories: (1) Unibiometric systems, and (2) Multibiometric systems.

Unibiometric systems

Traditionally, biometric recognition systems are unibiometric, which employ a single biometric trait for authentication purposes. It may use one of the physical or behavioral biometric traits, such as fingerprint recognition ( Maltoni et al., 2009 ; Hong, Wan & Jain, 1998 ; Yuan, Sun & Lv, 2016 ), face recognition ( Zhao et al., 2003 ; Masi et al., 2018 ; Wang et al., 2018 ), iris recognition ( Nguyen et al., 2017 ), signature, etc. In the literature, the use of unibiometric systems is widely employed with very good recognition results. However, such systems are typically constrained due to several factors including lack of accuracy due to noisy data, non-universality of biometric traits for registration, physiological limitations of biometrics, and vulnerabilities in biometric systems ( Table 1 ) ( Dinca & Hancke, 2017 ; Oloyede & Hancke, 2016 ).

Some biometric modalities are more vulnerable to some specific problems, e.g ., spoofing a fingerprint is relatively easier as compared to a vein/palm pattern. However, the recognition accuracy of fingerprints is far more superior. These are complementary properties of two different biometric modalities, which can be exploited together in a multimodal biometric system, hence making the system more tolerant to spoofing while maintaining a higher accuracy.

Multibiometric systems

When using the unibiometric systems, we may encounter problems due to several issues including, but not limited to, missing data ( Nandakumar, Jain & Ross, 2009 ) ( e.g ., occlusion in face image), poor sampling ( Grother & Tabassi, 2007 ), biometric duplication ( Sudhish, Jain & Cao, 2016 ), low discrimination among samples ( e.g ., hand shape/geometry) between distinct users, vulnerability to attacks, and spoofing, etc ( Jain & Kant, 2015 ). In situations like these, it may be necessary to make use of multiple biometric cues to boost the accuracy of a recognition system. The multibiometric systems offer so many features, making them more convenient and feasible as compared to the unimodal systems. There can be different sources of biometric information in a multibiometric system due to which such systems can be classified into the following five major categories ( Fig. 3 ):

Multi-sensor systems

The multi-sensor systems use multiple sensors in order to capture the same biometric trait of an individual ( Elhoseny et al., 2017 ; Kaur & Sohal, 2017 ). Such systems are desirable due to the fact that they can enhance the recognition capabilities of the systems ( Blum & Liu, 2005 ). This happens because the data acquired from various sensors may be of different quality and the multi-sensor system can partially solve the problems related to poor data quality ( Singh, Singh & Ross, 2019 ). The different biometric traits, when used for recognition, have the ability to complement one another, creating the possibility for a better recognition of the individuals.

Multi-algorithm systems

In multi-algorithm systems, more than one algorithm are utilized to improve the recognition rates of biometric systems ( Sotonwa & Oyeniran, 2019 ; Gad et al., 2018 ). It is cost effective to work on such systems as they do not make use of multiple biometric traits, and thus do not require multiple sensors ( Mishra, 2010 ). However, such systems require a lot of computational resources as multiple algorithms have to be run in order to calculate the relevant features for a single instance ( Fan et al., 2019 ; Scherhag, Rathgeb & Busch, 2018 ). Keeping this in view, special consideration should be given to the fact that real-time performance is a requirement of biometric systems, and thus the feasibility of such systems might be compromised even when they have the ability to achieve very high recognition rates.

Multi-sample systems

In multi-sample systems, multiple samples from the same sensor are acquired from the biometric devices ( Dinca & Hancke, 2017 ; Elhoseny et al., 2018 ). The fundamental issue with a single sample system is the fact that the samples can suffer from missing data problems due to which effective recognition cannot be performed ( Goswami et al., 2017 ). The problem is mitigated in multi-sample systems by acquiring multiple samples from the devices and using multiple or the most relevant samples for recognition. The same algorithm is used to process all samples and recognition results from each sample are calculated and eventually fused to yield a final result of recognition ( Modak & Jha, 2019 ). This recognition may be based on some technical considerations, e.g ., a confidence score with which a specific recognition result is obtained.

Multi-instance systems

In multi-instance systems, the biometric data is typically extracted from multiple instance of the same body traits ( Faltemier, Bowyer & Flynn, 2008 ). For example, finger biometric properties can be extracted from two fingers ( Lamia & Najoua, 2019 ), the palm prints can be acquired from two palms ( Leng et al., 2017 ), and the iris of the individuals ( Kumar, Prasad & Raju, 2020 ) can be used for measuring different biometric traits of the systems. The addition of multiple instances for performing recognition in a biometric system increases the discrimination capability of the system because the distinctive capability for a single individual is extended by adding more features to the pool, potentially leading to an improvement in the recognition rates for a system ( Leng et al., 2017 ).

Multi-modal systems

In the multi-modal systems, the biometric traits from different modalities can be combined together for the purpose of identification of an individual ( Modak & Jha, 2019 ). Such systems are used to complement the weaknesses of a single biometric and they usually try to make the best of different biometric traits in order to perform recognition of an individual ( Yang et al., 2018b ). An additional advantage of using multi-modal biometric systems is that they are more secure as compared to the uni-biometric systems as more than one trait is used at the time of registration of a user in a system ( Yang et al., 2018b ; Barni et al., 2019 ; Gomez-Barrero et al., 2017 ). Appropriately, stealing or forging one biometric trait does not guarantee an access to the system, thus leading to an improved security feature for authentication in biometric systems.

Designing a multibiometric system has a very high significance. A valid design will be able to ensure that the pieces of evidence collected from various sources, when fused together using different fusion strategies, can improve the recognition rates while ensuring some value added services provided to the users. However, when different modalities have to be combined to implement multibiometric systems, special consideration has to be given to several dimensions, e.g ., what kind of additional sensors will be required, what are the costs, is there a possibility to embed different sensors in the device and, what is the overhead of such a system in terms of computational complexity.

Performance metrics for evaluation

Multiple metrics can be employed to assess the performance of a biometric authentication system. Choosing a particular metric(s) depends upon the nature of evaluation. Following are the basic raw metrics and their descriptions:

• True Accept (TA) : A genuine user is correctly verified to its corresponding template stored within the biometric system.
• True Reject (TR ): An imposter is correctly rejected as its data does not match any template stored within the biometric system.
• False Accept (FA) : An imposter is incorrectly verified as a genuine user as his data is matched to the template stored within the biometric system.
• False Reject (FR) : A genuine user is incorrectly rejected as his data does not match any template stored within the biometric system.

The standard metrics that have been used to evaluate the performance of the authentication system in the literature are as follows:

• Equal Error Rate (EER) : Describes the point at which FAR and FRR are equal. Smaller values of EER refers to improved performance of a biometric system.

Fusion methods

Fusion plays a very considerable role in the implementation of multibiometric systems. There is an inherent requirement to fuse the information collected from different modalities before using it for the purpose of recognition. Fusion can be applied in multibiometric systems in two major settings: before matching and after matching. Consequently, there are five distinct levels at which fusion can be applied, i.e ., sensor level, score level, feature level, and decision level ( Fig. 4 ). The fusion applied at the first two levels is referred to as pre-matching fusion, whereas the rest are categorised as post-matching fusion.

Sensor level fusion

In sensor level fusion, the raw information is gathered from various sensors and is fused at the initial level prior to feature extraction to produce a raw fused information. Fusion of two images can take place at pixel, feature or at signal level. Fusion at sensor level can be between multiple samples of the same biometric gathered from multiple sensors ( Yang et al., 2005 ) or multiple instances of the same biometric taken from a single sensor ( Othman & Ross, 2012 ). Relatively less research has been done on this type of fusion in biometrics.

Feature level fusion

In feature level fusion, the features extracted from multiple biometric sources are combined together in the form of a single feature vector. In this fusion technique, features from different sensors, samples, and traits can be combined together. At this level of fusion, signals from various biometric channels are firstly pre-processed and their feature vectors are calculated independently; by using fusion algorithms, the feature vectors are fused to form a combined feature vector, which is used for recognition ( Xin et al., 2018 ). The incorporation of multimodal biometric traits in this type of fusion can be employed to exploit specific strengths of different biometric modalities ( Nagar, Nandakumar & Jain, 2011 ; Jagadiswary & Saraswady, 2016 ; Prasanalakshmi, Kannammal & Sridevi, 2011 ). Although better recognition results can be expected using this type of fusion technique, it has certain limitations including the lack of compatibility of different biometric features, curse of dimensionality, etc.

Matching score level fusion

This type of fusion is done by joining the scores yielded by the matching module of each feature vector with the template. The features are processed independently along with the calculation of scores, followed by the calculation of composite matching scores ( He et al., 2010 ; Yilmaz & Yankoglu, 2016 ; Kabir, Ahmad & Swamy, 2018 ). This is done by the checking the confidence scores, which are obtained using each feature vector. This type of fusion technique is typically easy and thus is being used by different multibiometric systems for effective execution.

Rank level fusion

In this type of fusion, sensor data is acquired followed by the feature extraction. The matching of this feature vector is performed against all the available templates in the database and similarity scores are obtained ( Kumar & Shekhar, 2010 ; Monwar & Gavrilova, 2009 ). The scores are arranged in the descending order and the entry corresponding to the lowest rank (indicating similarity of feature vector with the respective template) is taken as the most relevant to the data that is acquired from the sensor. The rank level fusion can also be employed for multibiometric traits and thus can yield a recognition score with a higher confidence. However, it should be noted that in addition to the pre-processing of sensor data, additional computational load is transfered on the matching module. Therefore, the rank level fusion can be computationally very complex especially when more than one biometric trait is employed.

Decision level fusion

In this type of fusion, the information obtained from different decision modules is combined together to decide about the identity of a user ( Jiang et al., 2014 ; Niu et al., 2008 ). The recognition results of each biometric trait are individually obtained followed by a fusion of these decisions to obtain a final decision regarding recognition ( Li et al., 2018 ; Ghosh, Sharma & Joshi, 2014 ). Various methods to perform this type of fusion can be used, e.g ., majority voting can be employed ( Jimenez, Morales-Morell & Creus, 1999 ). In systems which require enhanced security and fail safe functioning, rule based decision can also be made such as the use of a logical ‘AND’ operation, indicating that it is necessary for all biometric traits to be yielding the same output.

Hand based modalities

As discussed previously, there are many modalities that can be used to obtain biometric information from the users. Making the right choice for designing biometric systems is a question that requires consideration in multiple dimensions. The ease of use, budget, overall performance in terms of recognition ability and modalities that promote anti-spoofing are important influential factors that determine the best biometric trait for biometric security research. A brief overview of most of the physiological biometric modalities is summarized in Table 2 . The choice of the modality presents a trade-off between different factors, which require a careful review based on the nature of intended applications. As can be seen in Table 2 , retina and iris present the technologies which show very good recognition results. Physiologically, these modalities are highly distinctive for different individuals with almost no chance of repetition of the patterns. However, they are not user friendly, are expensive with respect to technology and highly sensitive to the protocols used for acquisition of the data. These specific limitations have resulted in a highly restrictive use of these two modalities, specially from the perspective of the convenience of the end user. Ear is one of the most stable biometric; however, it is not distinctive and is also sensitive to some external factors such as wearing of cap, jewelry, etc. Face is a physiologically motivated biometric and is very useful; however, the most fundamental flaw with the face biometric is that it is a source of infringement of the user’s privacy. Therefore, the users are typically not comfortable with hosting of their facial data specially by the third parties.

In contrast to the above-mentioned modalities, the remaining four, i.e ., fingerprint, palm print, hand vein, and hand geometry are the modalities which are based on hands. It should be noted that all these modalities (except hand geometry) are highly accurate for recognition, make use of inexpensive technology in general, are fast for matching as they require templates of very small size, and are less sensitive to acquisition conditions as compared to the other modalities that are used for biometrics. Therefore, in this paper, we have focused on a thorough review of the available literature on the use of biometrics based on hands ( Fig. 5 ).

Fingerprint

Fingerprint is one of the most established biometric modalities due to its high recognition rates and consistency, and has been in existence for over a century. The ease to acquire fingerprints and their wide usage has led to many commercial applications relying on them as far as biometrics are concerned. A fingerprint is formed by the coexistence of a collection of ridges and valleys, thus yielding a pattern, which is distinct for different human beings. These patterns are also referred to as “minutiae” and are mainly composed of bifurcations, enclosures, ridge endings and ridge dots. Further, the minutiae are subdivided into sub minutiae such as pores, crossovers, and deltas ( Fig. 6 ). A fingerprint biometric system has four main stages: acquisition of data, feature extraction, template creation, and matching. The ease of use and a small space required for the storage of template has made it one of the best biometric technologies to employ commercially.

On fingerprint biometric, both quantitative and qualitative works exist. A survey of around 160 users was done in Arteaga-Falconi, Al Osman & El Saddik (2015) and Cappelli et al. (2007) in which the users gave a positive response towards using this technology for smartphones. Furthermore, various technological contributions presenting quantitative results show that a fingerprint take less than one second for matching, achieve 0.07% EER on a database of 100 subjects, false rejection rates of upto 0.04% and false acceptance rates of upto 0%, 4.18% and 8.91% using three confidence coefficients, i.e ., 99.0%, 99.5% and 99.9% respectively. These results indicate a very high recognition performance in a very small amount of time, promoting the use of fingerprint technology for real time implementation of systems requiring biometric validation.

Palm print is a popular biometric modality, which has attracted the attention of many researchers. However, it is a relatively new biometric modality as compared to its counterparts, such as face, fingerprints, etc. A palm print image consists of some rich intrinsic features such as ridges and palm lines, delta lines, principle lines, minutiae features, wrinkles, etc. ( Chen, Huang & Zhou, 2013 ; Huang, Jia & Zhang, 2008 ) that are deemed to be permanent and unique for every individual ( Fig. 7 ). Owing to these inherent features, palm prints generate unique biometric characteristics for every individual that are reliable for identification purposes ( Zhang, Zuo & Yue, 2012 ; Zhong, Du & Zhong, 2019 ). The main issue that is responsible for reducing the performance of palm print systems is the deformation of images during the image acquisition process. Attempts are being made to solve this issue by using contact devices, but researchers have faced several challenges in the design of such devices including its size and limited usability, along with several challenges including position, stretch and rotation of the palm print. Lately, researchers are resorting to contactless devices again with low resolution imagery used for commercial application and high resolution imagery for applications such as criminal investigation.

Some of the most recent contributions on palm print biometric ( Zhang et al., 2017 ; Tabejamaat & Mousavi, 2017 ) show that the recognition rates of upto 98.78% and 97.2% respectively have been achieved within processing times in the order of milliseconds on databases of fairly large size (about 12,000 instances). The most relevant contribution in this regard with the best accuracies are presented by Luo et al. (2016) in which the authors have reported an accuracy of 100% on a dataset having 4,600 instances of palm prints. This is in contrast to a general perception that palm prints are not as accurate as fingerprints. However, it should be noted that the possibility to obtain a relatively large dataset to validate the findings on fingerprints having a statistically higher significance is much more likely in contrast to palm prints for which the availability of dataset is relatively limited.

Vein biometrics, also known as vascular biometrics, refer to a biometric system that measures parts of an individual’s circulatory system for identification. Vein pattern recognition technology has gained a significant attention due its unique attributes along with liveness property yielding very high recognition rates. Vein patterns are segmented into different sub-modalities amongst them most commonly used come from the palm ( Zhou & Kumar, 2011 ), palm dorsal ( Joardar, Chatterjee & Rakshit, 2014 ), wrist ( Pascual et al., 2010 ), or finger ( Lee et al., 2010 ). The sub-dermal nature of veins makes these types of biometrics a highly secure modality ( Crisan, 2017 ). In a vein biometric system, image acquisition is carried out by using near-infrared (NIR) imaging device. The NIR light maps the vein locations, because the hemoglobin in veins absorbs the NIR light. A high contrast image is created by visualization of the vein pattern as shadows appearing over a white background. These high contrast images with vein patterns are used for recognition using various texture feature extraction techniques.

A quick review of the literature elucidates the facts that very high recognition rates are obtained on this biometric trait. Researchers have reported an accuracy of upto 99.4% reinforcing the theoretical claims of high uniqueness of vein patterns ( Das et al., 2018 ). However, the requirement of using sophisticated acquisition devices for obtaining the biometric data makes this modality relatively less popular ( Kilian et al., 2020 ). Moreover, using vein patterns can be a challenge in some cases because of the physiological changes taking place due to ageing and various medical conditions ( Oloyede & Hancke, 2016 ).

Hand geometry

Hand geometry/shape is a very simple biometric technology that uses the measurements of human hand to verify the identity of the individuals. The measurements include the length, shape and width of fingers and size of palm ( Fig. 8 ). The biometric systems employing hand geometry are widely used as they have a high public acceptance ( Babich, 2012 ; Jain, Dass & Nandakumar, 2004 , Sharma et al., 2015 ). However, it should be noted that the systems based on this technology are not scalable as the hand geometry is not highly unique ( Oloyede & Hancke, 2016 ). Nevertheless, it is widely used at places providing access control, where the main objective is to find out if someone is illegally trying to gain access to someone’s personal identification. A hand reader guarantees that a worker is actually available at a place where he is meant to be. It is also used for implementing time attendance of the employees and helps in stopping the employees from buddy punching (which takes place commonly with fingerprint technology). Hence, the payroll accuracy of a company is guaranteed with a higher probability when hand geometry is used ( Babich, 2012 ).

Due to the lack of ability to differentiate between the people effectively, the usage of hand geometry is somewhat limited and typically used in conjunction with other biometric modalities for improved recognition rates. Some recent contributions on hand geometry show that an EER of upto 0.31% has been achieved by Sharma et al. (2015) with upto 50 distinct users. A novel contactless sensing system ( Kanhangad, Kumar & Zhang, 2011 ) based on multi sampling has been proposed, which has been used to authenticate a database of 100 people representing upto 200 hands with about 50% improvement in the recognition rates ( Oloyede & Hancke, 2016 ). Nevertheless, the technology is not as accurate as its counterparts and thus is not very useful in a standalone setting for large scale deployment for commercial purposes.

Finger knuckle print/inner knuckle print (FKP)

Finger knuckle print is one of the emerging hand based modalities used for biometric verification of the individuals ( Kumar & Ravikanth, 2009 ). The finger knuckle patterns can be easily acquired using contactless devices. In contrast to the more established modalities such as fingerprints requiring high resolution imagery, the knuckle patterns can be easily captured using low-resolution samples ( Zhang, Lu & Zhang, 2018 ). Additionally, the patterns on the outer surface of the knuckle appear at an early stage and survive for a longer periods of time and are specifically useful for the workers, labourers, cultivators, etc., whose fingerprints are more susceptible to damage due to the nature of work ( Yang, Yu & Liao, 2009 ). In a biometric system based on finger knuckle, the physiology which differentiates two different people is due to the lines, creases and texture of the knuckle print that lie at the three knuckle joints of the fingers ( Jaswal, Kaul & Nath, 2016 ) ( Fig. 9 ). These lines appear before birth and rarely change over an individual’s lifetime.

The knuckle print is a biometric that can be acquired without any physical contact with any sensor. Therefore, the chances of spoofing are significantly reduced. These are highly stable for individuals from various age groups; however, their widespread usage is still not reported. A quick survey of the literature shows that researchers have obtained a high accuracy on the identification of persons using knuckle prints with an overall accuracy of upto 98% in real time on a dataset of size 7,900; FAR of 0.062% and FRR of 0%. Given the ability to obtain data for finger knuckles contactlessly, the ease in acquisition process, invariance of patterns to emotions and behavioral aspects, and a wide acceptability socially, there is potential in using this technology on a large scale undoubtedly. However, research is needed on improving the identification results for a widespread deployment.

Feature extraction techniques for hand multibiometric systems

A significant volume of literature exists with regards to the design of hand multibiometric systems. It is well known that for any system, the feature extraction methodology is the most fundamental aspect that governs the recognition performance achieved by the systems. In this context, a survey of the literature reveals that all the methods can be mainly divided into the following main categories: 1. Statistical features, 2. Filter based features, 3. Deep features and 4. Hand crafted features, leading to a taxonomy, which is shown in Fig. 10 .

Statistical features

Texture is a spatial property that is generally valid over a larger spatial neighborhood. In order to capture the spatial characteristics and dependence of the images, some statistical measures can be used in order to summarize the patterns indicated in the images. Using statistical methods typically yield good quality image descriptors with the requirement of a smaller memory footprint for maintaining the templates of the registered users. Various authors have worked on proposing statistical methods for feature extraction (Summary presented in Table 3 ).

Aoyama, Ito & Aoki (2014) proposed a FKP recognition algorithm based on block matching using phase correlation where the phase information was extracted from the 2D DFT (Discrete Fourier Transform) of the images. Good recognition results were obtained with a high main-to-side lobe ratio of correlation and an EER of 0.321%. Perumal & Ramachandran (2015) proposed a method to fuse the palm print and FKP of individuals using interest point based techniques including SIFT (Scale Invariant Feature Transform), SURF (Speeded-up Robust Features) and EMD (Emiprical Mode Decomposition) with even better results but it should be noted that EMD is an iterative algorithm and is not very suitable for real-time implementations. Veluchamy & Karlmarx (2016) made use of FKP and finger veins to design a biometric system in which repeated line tracking is performed followed by grid operation, yielding a set of features, which are used for classification using support vector machines (SVM). Although the overall results are good, there is a margin for improvement before the method can be deployed for commercial usage. Yang & Sun (2016) proposed a biometric system making use of palm print and palm veins, in which local binary patterns were used for feature extraction. The method achieves good results; however, a wider validation on a larger dataset is required for a thorough evaluation of the proposed method. Srivastava et al. (2016) performed the fusion of palm-phalages, palm print and dorsal hand vein using some statistical features (average absolute deviation, mean features and Gaussian membership) followed by classification using SVM, KNN (k-nearest neighbors) and random forest. However, the main aim of the authors was to introduce a new dataset for carrying out biometric research.

Chaudhary, Srivastava & Bhardwaj (2016) used palm print and dorsal hand vein for recognition, performing feature extraction using Gaussian membership function. Bhilare et al. (2018) made use of the images from hand vein based modalities by performing an ROI extraction, followed by the use of CS-LBP (Center Symmetric Local Binary Patterns) as texture features. They obtained very good results, concluding in their research that the palm vein and finger vein images yield better results rather than using the vein structure of the entire hand for recognition. Vishi & Mavroeidis (2018) made use of fingerprint and finger vein for performing recognition. The paper was aimed at a combination of score normalization and fusion techniques on the two aforementioned modalities, concluding that the hyperbolic tangent score normalization technique achieves the highest recognition rates. Yang et al. (2018b) aimed at improving the template protection in a multibiometric system in which fingerprint and finger veins were used. They made use of DFT based features, achieving very good recognition rates while ensuring the security of the templates. Korichi et al. (2018) proposed a multibiometric identification system with modalities composed of images obtained from both visible and near-infrared light. Feature extraction is performed using PCA (principal component analysis), with results obtained on various datasets, showing a high precision.

Yang et al. (2018a) proposed a Weber representation based feature extraction method for feature extraction from finger veins and dorsal veins. They performed feature level fusion and validated their results on several datasets, achieving very good results. Zhang et al. (2019) made use of multiple finger based modalities, followed by feature extraction using graph based methods. They achieved very good results, showing a high potential of employing graph theory for designing biometric recognition systems. Also, the results are validated on a large dataset, elucidating on the statistical significance of the achieved performance. Veluchamy & Karlmarx (2020) proposed a multibiometric system based on FKP and palm print in which they used HoG (histogram of oriented gradients) features. These are also part of the MPEG-7 standard and thus are used widely for multimedia application. Good recognition rates have been achieved; however, there is a margin of improvement in the overall recognition rates that could possibly be achieved using more specialized descriptors. Lv et al. (2020) made use of LBP based image descriptor for feature extraction from the fingerprint and finger vein images. Feature level fusion was employed to combine the fingerprint and finger vein patterns into a single image, followed by the use of contrast enhancement techniques, and later performing LBP giving good results. You & Wang (2019) discussed about the classical disadvantage of the fuzzy vault scheme in terms of potential attacks carried out by the attackers. They mitigated this problem by performing the fusion of fingerprint and finger vein templates, followed by the projection of feature points on a rectangular grid. The fuzzy vault scheme is later used for encoding and decoding purposes. Good results are obtained along with experimental evidence of the security of the proposed scheme.

Filter based features

Several feature extraction methods exist in the literature that are based on first filtering the images with a specific mask (filter) or a set of filters and then estimating the texture of the images based on models or statistics of the filter outputs. Typically, these methods result in a decomposition of the images giving a large amount of data based on the number of filters and their parameters (more descriptive with respect to some texture related characteristics such as edges, scales, angles, etc). Later, the decomposed information is summarized using some statistics, which enhance the description of the images with respect to the filter parameters. The literature is rich in terms of the use of filter based methods for texture feature extraction (Summary in Table 4 ).

Chin et al. (2014) predominantly focus on multibiometirc template protection. They have proposed a three-stage hybrid method. In order to obtain a fused template, fusion of fingerprint and palm print images is done at feature level, followed by applying random tile technique to obtain random features. These random fused features undergo discretisation, hence generating a secure template bit string. Khellat-Kihel et al. (2016) pointed out that the multimodal biometric system improves the accuracy significantly but on the other hand, they tend to have a larger memory footprint and result in longer execution times. Therefore, they proposed the extraction of features using Gabor filters, followed by feature selection using linear discriminant analysis (LDA) giving good recognition results. Gupta, Srivastava & Gupta (2016) proposed a hand geometry and vein pattern based method in which gradient based variational approach is used for the extraction of veins. Matching is performed using global approach in which Fourier Mellin transform is used, thus avoiding the issues such as non-uniform illumination, noise, etc. Hand geometry features are obtained using hand crafted features. The authors have carried out validation of methods on a large dataset with statistically significant results.

Verma & Dubey (2017) proposed a multimodal vein based recognition system in which an ROI (region of interest) is identified followed by vein enhancement using circular averaging filter and holoentropy thresholding. The results reported by Verma & Dubey (2017) are not very good; however, the feature extraction procedure is not precisely mentioned, and thus the quality of features employed cannot be fairly assessed. Bharathi & Sudhakar (2019) made use of hand vein based biometric modalities for performing recognition. Feature extraction was performed using Gabor filter and gradient based methods with matching performed using the Euclidean distance metric. Although good results are obtained, the authors have mentioned that using some other fusion techniques can improve the results. Furthermore, researching on a more relevant distance metric could potentially be useful for matching purposes. Kauba, Prommegger & Uhl (2019) proposed a contactless device to acquire images corresponding to hand-based biometric modalities, which make use of vein patterns for recognition. The authors collected a dataset for evaluation and also used various methods including Gabor filter, high frequency filtering, and interest point based methods for feature extraction. The proposed device achieves good results and exhibits potential for usage for recognition tasks. Jaswal & Poonia (2020) made use of palm print and finger knuckles to design an authentication system. They performed an ROI extraction from the respective images followed by line ordinal pattern based encoding of the images. Later, feature extraction is performed using criterion wavelet transform and feature selection is performed using linear discriminant analysis and search based methods yielding very good recognition rates. Li et al. (2021) proposed a joint discriminative feature learning framework in which the directional features are estimated using Gabor filters, which are later fed into an optimization framework for feature learning that maximizes the inter-class variation and minimizes the intra-class variation among samples. Finally, block-wise histograms of learned feature maps are used for recognition purposes, giving very good overall recognition accuracy of about 99.65%.

Deep features

Over the last couple of years, the Deep Convolutional Neural Networks have dominated significantly in terms of the extraction of features and for performing the classification tasks or solving recognition problems. This is because of their robust framework, having an incredible ability to learn from the training data and adapt the designed networks to solve complex problems. Even in biometric systems, the employment of deep learning has seen a significant surge and is producing very good results in comparison to the other methods that have been used previously (Summary in Table 5 ).

Zhong et al. (2018) proposed the use of DHN (deep hashing network) for palm print encoding into 128-bit codes, and BGM (biometric graph matching) to encode dorsal hand vein images into three discriminant features. Later, feature level fusion was used with very good recognition rates, with EER of upto 0%. Toygar, Babalola & Bitrim (2020) proposed a deep architecture with five hidden layers, each comprising convolutional, batch normalization and pooling layers to design a multibiometric system based on palm, dorsal, and wrist veins. The results when compared with several other methods including hand crafted features and Gabor filters, show very good results, elucidating on the potential of using deep learning methods for multimodal biometrics. Zhong, Shao & Du (2019) proposed a deep end-to-end trainable hashing network that takes an image at the input and outputs a binary code corresponding to the respective image. Matching can be performed by comparing the binary codes corresponding to the training images with the image given as input to the network. The method achieves very good results, showing promise in employing the neural networks based techniques for encoding the images. Chen et al. (2019) introduced a low cost personal identification system consisting of near infrared and visible LED (light emitting diodes). An adaptive feedback control was used to control the brightness of the diodes. The images acquired were preprocessed, with feature extraction performed using a deep scattering CNN, giving good recognition rates. Mehdi Cherrat, Alaoui & Bouzahir (2020) proposed a system for recognition using the CNN models in a multibiometric setting with a fusion of finger vein and fingerprint. Good recognition rates were obtained using the proposed strategy, with a conclusion that the use of preprocessing improves the recognition rates. Choudhury, Kumar & Laskar (2021) made use of index, middle, and ring fingernail plates for extracting biometric features from the images using three customized pretrained models: Alexnet, Resnet and Densenet. An adaptive fusion technique based on score and decision level is used for the purpose of fusion of features from dorsal hand followed by exhaustive experiments to assess the efficacy of the proposed technique, giving very good recognition results with a minimum average rate of 0.0097%.

Hand crafted features

Handcrafted features are typically those which are used with more traditional machine learning algorithms for performing classification tasks. More commonly, these features can be easily correlated to statistical features. In this specific article, we define the handcrafted features as those features that are obtained as a combination of statistical and physical properties of the images such as hand size, finger size, etc. Such combinations are typically obtained when at least one of the biometric traits used for recognition is hand geometry. Although the literature on such techniques is limited, a summary of relevant contributions made using such methods is presented in Table 6 .

Sharma et al. (2015) performed identity verification using the shape and geometry of hands using the contour of the hands. The hand contour is initially aligned followed by the calculation of peaks and valleys, and the extraction of the finger feature points by calculating the Euclidean distance between the reference point and all the feature points. The proposed method shows promise with good recognition rates achieved over two datasets. However, their sizes are small and the methodology requires extensive validation over larger datasets for a statistically significant conclusion. Anitha & Rao (2016) made use of FKP and hand geometry to propose a multibiometric system. They performed ROI extraction from the pictures followed by the use of LBP as texture features for the finger knuckles and hand geometry features using hand crafted features. The process for the extraction of these features includes the identification of six points on the hand to extract the angle features and the aspect ratio of the palm, followed by the calculation of Euclidean distance between the template and the acquired image for matching. Experiments show that the best performance is achieved using a feature level fusion of the FKP and hand geometry features. Jaswal et al. (2019) laid down the idea of using multiple biometric traits for recognition using a single sensor. A device having the ability to capture the FKP and palm print was used for acquiring the image. Processing of the images was done by the extraction of an ROI followed by a transformation using texture code matrix. Hand registration was done by detecting the feature points (peaks and valleys) and also the detection of knuckle point followed by deep multiscale matching giving very good recognition rates. Gupta & Gupta (2018) proposed a system that captures slap fingerprints and hand dorsal image at the same time. Slap segmentation is performed by making use of the finger location and hand type. Matching of scores are generated by matching the slap fingerprints, palm dorsal vein and hand geometry that are fused for the purpose of authentication yielding good recognition results. Khodadoust et al. (2021) worked on the fusion of fingerprint, finger veins, and FKP using an experimental setup, which obtained the 3D reconstruction of the above-mentioned traits followed by maximum curvature based feature extraction. They obtained good overall identification results, with validation carried out on 66 users showing promise for the use of their method for authentication purposes. The most significant claimed advantage of the proposed method relates to the experimental protocol as the method relies on a contactless and hygienic way of acquiring multibiometric traits.

Now that we have analyzed the feature extraction techniques, which have been presented in the literature, it is important to note that there are a several points to compromise a biometric system. It is very important for a biometric system to be unsusceptible to attacks and loss of template by adversaries. To deal with the issues related to the security of a multibiometric template, we now analyze the existing work on multibiometric template protection.

Multibiometric template security

A multibiometric system uses multiple biometric traits ( e.g ., fingerprint, face, and finger vein) to recognize a person ( Ross, Nandakumar & Jain, 2008 ), hence improving the reliability and accuracy of biometric systems. However, adequate attention has not been paid towards making the multibiometric templates secure. There are several ways to compromise a biometric system ( Ratha, Connell & Bolle, 2001 ) and loss of a biometric template information to unauthorized individuals possesses security and privacy threats ( Nagar, Nandakumar & Jain, 2011 ; Mirza et al., 2014 ) due to following reasons:

• Intrusion attack at biometric system: If an adversary gets an unauthorised access into a biometric system, he can easily access the stored biometric template of a user. This information can be used to get an illegal entrance into the biometric system in which the user is enrolled by either reverse engineering the template and disguising as this user or replaying the stolen template.
• Database Linkage : Once an adversary gets hold of a template, it can be easily determined if the two templates from different databases belong to the same person or not. Moreover, different databases hold separate parts of data regarding that person. Consequently leading to more data theft and more difficult identity-related attacks.

Keeping this in view, the security of a multibiometric system is very critical as it contains information regarding multiple biometric traits of the same user and it should be shielded from an unauthorized access ( Chin et al., 2014 ; Rathgeb & Busch, 2012 ). Therefore, there is a need for a secure template that must be irreversible and unlinkable ( Davida, Frankel & Matt, 1998 ; Ratha et al., 2007 ; Bolle, Connell & Ratha, 2002 ; Juels & Sudan, 2006 ; Sutcu, Li & Memon, 2007a , Sarkar & Singh, 2020 ; Bharathi & Mohana, 2019 ) ( Fig. 11 ). Biometric template protection schemes can be categorized into two main classes ( Fig. 12 ) ( Rathgeb & Uhl, 2011 ; Sandhya & Prasad, 2017 ): 1. Cancelable Biometrics (CB), 2. Biometric Cryptosystems (BCs). These schemes offer various advantages over a generic biometric systems. A few most important advantages are summarized in Table 7 .

Cancelable biometrics

Cancelable biometrics (CB) refer to distortion of biometric features that are intentional and systematically repeatable in nature to protect sensitive user-specific data ( Ratha, Connell & Bolle, 2001 ). Cancelable biometric transforms are those that are used to transform the original biometric samples such that the resulting data is computationally hard to recover. When the user registers in the system, his biometric sample is transformed using a one-way transformation and saved in the database. This transformation is chosen from an identification word that is specific to the user. In the verification step, the query template is used to generate a transformed template that is compared with the saved template in the database followed by the verification process. The literature on unimodal cancelable biometric systems is very rich but there are inherent problems with such systems including intraclass variability, variation in data quality and a significant similarity in interclass samples. In contrast to such systems, the multimodal biometric systems combine the features of various biometric traits to generate the templates, which are more secure and thus, resistant to various threats and attacks. The main advantages offered by the multibiometric systems are greater security, accuracy, noise sensitivity, and resistance to spoof attacks. Table 8 presents a brief summary of the work done in the domain of cancelable multibiometircs.

Researchers have made several contributions on multibiometric template protection employing cancelable biometrics. Paul & Gavrilova (2012) proposed a method in which two-fold random selections are made from each biometric trait, followed by a feature level fusion. Random projection of each fold is obtained followed by PCA (principal component analysis) and later K-means clustering to generate the single template for individual biometrics. Later, LDA (linear discriminant analysis) is applied to further improve discriminability of the features. Final authentication is carried out using a classifier. Another variant of the technique proposed in Paul & Gavrilova (2013) makes use of Gram–Schmidt transformation instead of PCA, along with some other minor modifications in the pipeline. The authors have validated the cancelable property of the proposed method, while giving good authentication results in a multibiometric setting. Furthermore, the authors improved the results by proposing a methodology in which both Gram–Schmidt transformation and PCA were used followed by a rank level fusion for performing final authentication of the users ( Paul & Gavrilova, 2014 ). Chin et al. (2014) proposed a three-stage hybird template protection scheme. They have performed the fusion of palm print and fingerprint on the feature level, followed by the use of random tiling technique to extract unique features. Finally, the fused random features undergo 2 N discretisation to produce the template bit string. The approach addresses the criterion for template protection with an improved EER as compared to unimodal biometric sytems, though it is slightly higher with reference to multimodal systems. Gomez-Barrero et al. (2017) made use of homomorphic probabilistic encryption to generate the biometric templates along with fusion at three different levels. A complexity analysis was also carried out to assess the feasibility of the proposed method for real-time implementation. Moreover, feature level fusion is also employed yielding more secure and better cancelable biometric features. Kaur & Khanna (2018) proposed a template transformation method named random distance method that yields privacy preserving, revocable and discriminative pseudo biometric identities, with about 50% reduced memory footprints. Yang et al. (2018b) proposed a multibiometric system in which the fingerprint based minutae features and finger vein features are extracted followed by their respective binary features, and then performing feature level fusion in three different ways. The method obtained secure biometric templates with good recognition results. Gomez-Barrero et al. (2018) showed the use of Bloom filter based protection schemes while elucidating that it is not a straightforward task. A statistical analysis of unprotected templates is carried out to estimate the main parameters of such schemes. Dwivedi & Dey (2019) proposed a method to obtain cancelable templates by using log-Gabor filters with phase quantization, followed by the generation of biometric codes. Score level fusion from multiple biometric templates are used for authentication yielding better results in comparison to unibiometric systems with better accuracy. Walia et al. (2020) proposed a method to obtain cancelable features using deep neural networks that are fused using adaptive graph based fusion method. The proposed method is used to obtain multimodal unified templates, which are empirically demonstrated to be robust to adversary attacks. Chang et al. (2020) proposed an authentication approach in which bit-wise encryption scheme is used to transform a biometric template into a secure template using a secret key, which is generated from another template. The scheme fully preserves the number of bit errors in the protected and original template, ensuring that the recognition performance is the same as that in the case of unprotected templates.

Biometric cryptosystems

Biometric cryptosystems (BCs) refer to designs that securely generate digital key from a biometric or bind a digital key to a biometric ( Cavoukian & Stoianov, 2009 ). To overcome the shortcomings of traditional verification methods, which were based on password-based key-release, BCs bring about a considerable security benefit by offering biometric-dependant key-release since the biometrics have a strong link with the user’s identity ( Uludag et al., 2004 ; Jain, Ross & Uludag, 2005 ; Rathgeb & Uhl, 2011 ). At the same time, combining biometrics with cryptography and extracting the keys is not that straightforward due to variations present in a biometric data. Most of the BCs require helper data that contains additional information about the biometric and is used to generate or retrieve a key ( Jain, Nandakumar & Nagar, 2008 ; Ali & Khan, 2014 ). A helper data must not reveal significant information about original biometric templates. Table 9 presents a brief summary of the biometric cryptosystems.

Sutcu, Li & Memom (2007b) proposed the use of multibiometric features, followed by a secure sketch block, making it hard to extricate the original samples from the encrypted features. Nandakumar & Jain (2008) proposed a method to enable template protection using fuzzy vault framework. The authors claim to improve the recognition performance of the system along with enhanced security. Camlikaya, Kholmatov & Yanikoglu (2008) proposed a technique for the fusion of fingerprint template along with behavioral biometrics (voice samples). The algorithm enhanced the security of the biometric system by encoding the fingerprint features within the voice feature vector. The use of voice was motivated by using the property of spoken words used as a password to achieve the desirable cancelable property. Multiple biometric cryptosystems were proposed by Fu et al. (2009) out of which three were used for performing biometric fusion at the cryptographic level. The authors presented no experimental results; however, a detailed theoretical analysis of algorithms, comparison, and discussion were carried out. Nagar, Nandakumar & Jain (2011) provided a feature-level fusion method for both fuzzy vault and fuzzy commitment schemes that simultaneously secure the multiple templates of a user using a single secure sketch. Feature level fusion using multiple characteristics of a user proves to be significant in providing high privacy as compared to the single characteristic biometric systems, since only the fused feature vector is stored on the server database. Further, it requires less storage since only the combined feature vector is stored in the database server. However, it requires additional feature extraction and transformation tools for the heterogeneous features (variable formats based on distance, similarity, etc.). Another hybrid methodology to secure the biometric systems was proposed by Li et al. (2015) in which a combination of computational security and information security principles was implemented. Decision level fusion was done in the proposed cryptosystem for performing recognition. Kumar & Kumar (2015) proposed a multibiometric system based on cell array. Encoding and hash code computation was done using Bose Chaudhuri Hocquenghem (BCH) on the biometric modalities. The data is scattered across the two cells such that the first cell stores the hash code and the second cell stores the key. Moreover, fusion was performed at both decision and feature levels out of which the former shows better results in a multibiometric cryptosystem setting. You & Wang (2019) proposed a novel fuzzy vault scheme, which effectively protects the multibiometric template against location attack, brute force attack, and correlation attack. They performed fusion of fingerprint and finger vein templates. Feature point fusion encoding is done through grid projection, and fusion encodings are applied to construct the fuzzy vault. Chang et al. (2021) proposed BIOFUSE in which fuzzy commitment and fuzzy vault are combined using an encryption scheme. The system makes it difficult for an attacker to gain unauthorized access to the system without doing an impersonation of all the biometric traits at the same instant. The experiments have shown very good recognition rates with a high security. Evangelin & Fred (2021) used a visual shadow creation process to create multiple shadows of one image followed by encoding and decoding using elliptic curve cryptography. Although a very secure model is obtained, the implementation time of the model was significantly expanded. Asthana, Walia & Gupta (2021) made use of a key binding mechanism to generate a secret key using the biometric data of the user, leading to the proposition of a biocrypto system. Novel objective functions are proposed to create the helper data. The local minima of the objective function are taken as anchor points to retrieve the secret key and perform recognition leading to about 98% success rate in recognition even in the presence of limited noise in biometric data.

Open challenges and future directions

The limitations faced by various researchers in the implementation of biometric systems are listed as under:

• Aging/Alteration: It is well known that even when the biometric traits do not suffer any natural changes over a period of time, they are subject to changes due to trauma or physical damage due to cuts, different skin conditions, or other unforeseen events. Even when there is no medical/physical condition, which is liable to cause any damage to the biometrics, the traits change over a period of time ( Lanitis, 2010 ; Trokielewicz, Czajka & Maciejewicz, 2018 ). This is a challenging problem and as such there is no remedial solution available for such problems other than making use of the biometric traits that are less sensitive to such alterations. Among the hand biometric systems, there are several modalities which exhibit the information not from the skin but the sub-skin structures (veins) that are captured using the IR camera. A mild or superficial skin condition does not affect the vasculature, hence the vein based multibiometric systems have the ability to cater for most of the problems occurring due to trauma on the skin such as cuts. Aging and some other chronic conditions such as hypertension, diabetes, etc., affect the vein biometrics (changing in the diameter of the veins) and thus can impact on the recognition performance of the biometric systems ( Xie et al., 2017 ). However, since these changes do not take place overnight, there are ways to mitigate these problems.
• Operational problems: Operational problems refer to the various problems that have the ability to affect the performance of multibiometric systems. These problems can result from various factors such as environment and the methodology of acquisition of the data. The good thing is that most of these operational problems can be mitigated by the acquisition of data multiple times, until a good sample is captured from the acquisition device. For example, if there is excessive moisture in the fingers, the sample captured from the device may be having specular reflection ( Auksorius & Boccara, 2017 ). Such problems can be mitigated if the fingers/hands are cleaned for any moisture before application to the sensor. Apart from environmental conditions, even when the proper methodologies for acquisition of data have been followed, it is possible to face some issues such as alignment. If the biometrics are not properly aligned according to the templates and are, for example, captured at different angles than the templates, then it is possible to handle such problems with one of the modules, such as feature extraction. Therefore, it should be noted that the features should be extracted such that they are invariant to some underlying imaging conditions such as illumination and rotation at the very least.
• User errors: The orientation and shifting of the fingers during registration and authentication process significantly affects the performance of the biometric systems based on fingers ( Liu et al., 2014 ; Peng et al., 2012 ; Dong et al., 2014 ). Furthermore, during the imaging process, any movement of the fingers or hands causes irregular illumination ( Song et al., 2011 ). As a result, different segments of the fingers/hands get different amount of light absorption and as a result, the quality of the acquired images is not adequate for performing recognition.
• Biometric finger features: Another important factor which plays its role is finger features. Studies have shown that in addition to the varying thickness of fingers, certain factors related to finger skin affect the image capturing process, such as skin pigmentation, thickness, hair, etc. ( Gupta & Gupta, 2015 ). Furthermore, studies have shown that the varying thickness of finger skin results in an unequal distribution of light passing through the skin, hampering in the collection of high quality vein patterns.
• Complexity of fusion: It is clear that the use of multibiometrics leads to a better recognition and can help in increasing the security of the systems. However, with the use of more than one trait, there are four possibilities of performing the process of fusion on the biometric templates. This leads to critical decision making in any system as it could define the performance as well as the security characteristics of the systems ( Dinca & Hancke, 2017 ; Zhong, Du & Zhong, 2019 ). Moreover, biometrics is a domain that requires real-time performance and fusion will typically require the system to process a large amount of data in comparison to a unibiometric system (which does not require fusion). The recognition rates by adopting various methods of fusion could vary based on the platforms ( e.g ., mobile, wearable devices, etc.) and architectures (biometric traits, feature extraction methods). Therefore, an exploratory study of the performance yielded by different levels of fusion using standardized platforms and architectures to analyze the best fusion methods is a significant challenge.
• Security of multibiometric templates: The security aspect in biometric systems has recently gained a significant traction. There are several techniques that have been explored well including different fusion schemes, convolutional methods using different types of filters, methods for generating secure sketches and fuzzy vault constructs, encryption schemes, etc. In this context, the literature is very rich in regard to the unibiometric systems. The literature on the security of multibiometric systems is very limited and thus there is a lot of scope of work available in this area. The most widely explored methods used for multibiometric systems are derived from transformation based methods and fusion techniques. Generally, the fusion techniques are used in conjunction with a variety of other methods used for inducing security in multibiometric systems. This is because, there are different types of fusion that take place at different stages in the pipeline of multibiometric systems. However, it should be noted that the feature based fusion is the most common type of fusion technique, which is used for enhancing security, as it gives a richer set of features, generally yielding better evaluation metrics for multibiometric systems. It would not be wrong to say that most of the proposed techniques are focused on the use of hand crafted features, along with a mixture of fusion techniques, with a limited focus on representation learning based algorithms.
• Lack of standard performance measures: There are several measures that have been used to assess the performance of a biometric system. Due to the lack of a single measure to quantify the system, it is very hard to make a comparison of different methods that have been published in the literature ( Ryu et al., 2021 ; Kumar, Prasad & Raju, 2020 ; Manisha & Kumar, 2020 ). There is a need to work on a unified measure that is widely adopted by the researchers to evaluate different authentication systems, and making the literature standardized and valid for performing direct comparisons.

Future Directions for Improving Biometric Systems

The main future directions of work for the implementation of biometric systems are summarized as follows:

• Multi-sample registration: One method that is typically used to solve the problem of variations in the data acquisition for a single person is doing a multi-sample registration, in which multiple samples are captured for a specific trait. In this way, the variations of a single sample are captured and the machine learning methods are appropriately trained to capture this intra-sample variation.
• Rotation and illumination invariant descriptors: As discussed previously, one problem that is faced during the sample acquisition is capturing the samples at varying angles. It is practically not possible to always capture the data of the hand biometrics exactly aligned according to the available templates. This could be due to both the acquisition of the template or the sample. An interesting way to mitigate this problem is to work on image descriptors, which are invariant to the image capturing conditions ( Riaz et al., 2013 ). Specifically, if the descriptors are rotation and illumination invariant, this problem can be effectively addressed. However, it should be noted that if the descriptors are illumination invariant, it would not be possible to integrate the information about soft biometrics (such as color of hand) within the template and the respective features.
• Incremental machine learning (IML): A very interesting area that has recently gained traction is IML. As discussed, the biometrics are bound to changes over a period of time. Some biometric traits undergo more transformations as compared to the other. If somehow these changes are properly recorded, they can be effectively handled while still yielding higher recognition rates. There are two main methods to solve these issues ( Mehrotra et al., 2016 ): one way is to keep updating the templates as the user is authenticated. Another way is to make the ML algorithm learn new parameters while a new sample is presented for a user. Both these methods are effectively used for incorporating adaptivity to some extent.
• Fusion based adaptation: The process of adaptation in biometrics is not only limtied to templates, but it can also be extended to the fusion level. For instance, when performing fusion of multibiometrics, it is possible to perform fusion on the decision level by giving more weightage to the traits which are more stable over a period of time as compared to those which are not. In this context, there are several examples in the literature with regard to adaptive score weighting ( Assaad & Serpen, 2015 ; Sim et al., 2014 ), score normalization ( Khalifa, Gazzah & BenAmara, 2013 ), adaptive feature weighting ( Huang et al., 2015 ; Xu & Lu, 2015 ), etc.
• Soft biometrics: This paper mainly discusses the literature on hard biometrics in which the physiological biometric traits are used to authenticate the users based on their mathematical modeling. Lately; however, soft biometrics have been gaining attraction since they have the ability to complement the biometric systems with their decision making. The idea behind this concept is that the biometric systems make decision about a specific user based on some characteristics such has skin colour, eye color, height, weight, beard, etc. Interestingly, hand based biometric systems take this liberty to use these soft biometrics to perform recognition. This is because, there are at least two very important characteristics that can be used for recognition in hand based multibiometric systems, i.e ., skin color and hair. The skin color can be used as a property of the individuals, whereas the presence of hair on the hands or the texture of hands from dorsal view can indicate the gender of the individual. Also, previous studies have shown that the hand measurements, hand length, hand breadth, palm length, palm breadth, etc., can be correlated to the gender of an individual ( Rastogi, Murali & Rastogi, 2014 ). Soft biometrics can be used for recognition in a mixed authentication setting where these are used in conjunction with the hard biometrics for authentication.
• Multibiometric template protection schemes: Although the work on multibiometric template protection is limited, this is bound to change in the future given that there is an increasing interest of researchers in extending such algorithms with their variants that yield better results. Moreover, the focus of researchers currently is on using the deep features/algorithms in various domains rather than focusing on the classical classification algorithms ( Khan et al., 2019 ). A notable recent contribution related to cancelable biometrics using deep features is the technique proposed in ( Walia et al., 2020 ), where the authors have used a modified of Resnet model generating deep representation of biometric traits, followed by graph based fusion for generating a unified template. Given that cancelable biometrics make the biometric templates non-recoverable offering a high security, but the recognition rate is expected to be compromised. The method offers better results in comparison to some other methods considered in this paper; however, there is a scope for improvement as can be seen by the performance where the EER of the proposed method is 4.34%. Another potential direction for future research is the use of hybrid methods for template protection, which offers combined benefits of several methods. It should be noted that this requires caution due to the fact that the performance of authentication should still stay real-time and the employment of multiple computationally complex algorithms may require powerful computing resources.

Conclusions

In this article, we have performed a detailed survey of hand-based multibiometric systems. In this context, various hand-based biometric modalities are discussed, along with a through discussion about various fusion techniques employed and a brief survey of recent work that is being done on template protection schemes in biometric systems. A summary of the main conclusions is as follows:

• The acquisition of biometric templates is a process, which is controlled by the user and thus has the ability to incur some unexpected variations. This can be handled using invariant image descriptors; however, incremental ML is one area that can be explored to solve such problems.
• Lately, hard biometrics can be combined with soft biometrics for authentication purposes. The hand-based modalities give this liberty to extract soft biometrics that can lead to the improvement in biometric security and authentication.
• Most of the work done on the security of biometric templates is employed on unibiometric systems. Multibiometrics in the context of security is largely unexplored with significant margin of improvement for future contributions to the domain.
• Recently, ML algorithms are being used outside their conventional usage as classification tools and multibiometric security is one such area. The potential of deep learning is demonstrated in the literature, with a great margin for improvement.

Funding Statement

This project was funded by the National Plan for Science, Technology and Innovation (MAARIFAH), King Abdulaziz City for Science and Technology, Kingdom of Saudi Arabia, Award Number (3-17-09-001-0008). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Additional Information and Declarations

Farrukh Aslam Khan and Haider Abbas are Academic Editors for PeerJ.

Anum Aftab conceived and designed the experiments, analyzed the data, prepared figures and/or tables, and approved the final draft.

Farrukh Aslam Khan conceived and designed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the paper, mentoring student, and approved the final draft.

Muhammad Khurram Khan conceived and designed the experiments, analyzed the data, prepared figures and/or tables, mentoring student, and approved the final draft.

Haider Abbas conceived and designed the experiments, analyzed the data, prepared figures and/or tables, mentoring student, and approved the final draft.

Waseem Iqbal conceived and designed the experiments, analyzed the data, prepared figures and/or tables, and approved the final draft.

Farhan Riaz conceived and designed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the paper, and approved the final draft.

A Review of Face Recognition Technology

Ieee account.

• Change Username/Password
• Update Address

Purchase Details

• Payment Options
• Order History
• View Purchased Documents

Profile Information

• Communications Preferences
• Profession and Education
• Technical Interests
• US & Canada: +1 800 678 4333
• Worldwide: +1 732 981 0060
• Contact & Support
• About IEEE Xplore
• Accessibility
• Terms of Use
• Nondiscrimination Policy
• Privacy & Opting Out of Cookies

A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. © Copyright 2024 IEEE - All rights reserved. Use of this web site signifies your agreement to the terms and conditions.

Integrated Biometrics introduces most compact mobile palm scanner

Integrated Biometrics has launched what it calls an industry-first compact, mobile palm scanner designed for use in forensics, law enforcement, and special operations and identity applications globally. A release says Mannix uses IB’s patented Zero Bezel Film (ZTX) technology to enable intuitive palm and fingerprint capture on a FAP 60 FBI-certified device.

“What makes Mannix so unique from other scanners is that it is so small, lightweight and rugged, yet it can be powered simply by your mobile phone,” says David Gerulski, EVP at Integrated Biometrics. The scanner’s dimensions are 161 mm x 161 mm x 20 mm (6.3″ x 6.3″ x 0.8″), making it slightly larger than the average adult human hand. Since it has no bezel – the border between a mobile device’s screen and its frame – the entire surface is photosensitive, allowing IB’s patented LES film technology to capture upper palms, lower palms, writer’s palms, and rolled prints, as well as traditional fingerprints .

The scanner is also designed to perform in challenging conditions with resistance to dust, moisture and extreme temperatures, IB says.

In addition to the standalone model, the Mannix palm scanner is also being offered as an embedded module powered by IB’s SDK driving Android and Windows devices.

“Mannix is the answer to long-awaited market demand and is the perfect addition to IB’s hardware lineup,” says Shawn O’Rourke, CEO of Integrated Biometrics. “Now, with Zero Bezel film technology, it provides total flexibility for natural, intuitive fingerprint scanning.”

Related Posts

Article Topics

biometrics  |  fingerprint biometrics  |  Integrated Biometrics  |  Mannix  |  palm biometrics  |  palm scanner

Latest Biometrics News

Uk invests in 4 police vans with live facial recognition to cut knife crime.

The UK government has announced investments in live facial recognition and research and development for remote scanning technologies as it…

Passkeys continue march to mainstream with Visa, WhatsApp updates

Visa has unveiled new digital products and services based on biometrics and passkeys, as it aims to address rapid changes…

States should modernize identity verification to meet fraud challenge: Socure

U.S. state governments are facing major losses due to fraud, but the tools to improve the outcomes of digital interactions…

Digital identity protection tool backed by Australian bank sends alerts for fraud

The venture scaling arm of Australia’s biggest bank has launched a digital identity protection tool that will alert users to…

NADRA expands digital identity card services to post offices, trans people

The National Database and Registration Authority (NADRA) of Pakistan has expanded its digital identity services to enhance accessibility and inclusion,…

Nordic and Baltic countries need better cross-border digital identity matching: NOBID

Nordic and Baltic countries need better identity matching solutions to make cross-border digital identity a reality, concludes a new report…

Leave a Reply Cancel reply

This site uses Akismet to reduce spam. Learn how your comment data is processed .

Continue Reading

Most Read This Week

• Cameroon hands 15-year biometric ID card contract to Augentic
• Bahrain to overhaul ID card system with Thales contract
• Australia outlines new budget for digital ID scheme
• Peru’s digital ID under magnifying glass after national identification registry hit by criticism
• Malawi on path to inclusion with digital ID project supported by EU, UNDP
• Biometric smart cards deployed to new markets for improved payment, access security
• Face biometrics secure payments, borders, public services, but good practice eludes some
• European digital ID project offers grants for more pilots
• Paris rolls out AI surveillance for Taylor Swift
• Malaysia greenlights Worldcoin token trading, setting up biometric crypto growth
• Daily Biometrics News

Featured Company

• More Biometrics Companies

Biometrics Insight, Opinion

• Deepfake fears motivate consumers to change behavior, but misconceptions linger
• Redefining online safety: The evolution of age verification laws
• Startup OVE reveals prototype for flexible biometric payment device
• Video injection attacks: What is that and the way forward?
• Biometrics Industry Interviews

Digital ID In-Depth

• Sri Lanka’s digital identity transformation seeks balance of speed and assurance
• Court orders changes to Tanzania’s data protection law
• Pakistan’s NFIS initiative leverages digital identity for financial inclusion
• Governance framework key to trust that enables data-driven policymaking
• Digital ID Biometrics

Biometrics White Papers

• Protect your business and customers from threat of account fraud
• A buyer’s guide to online identity verification
• Generative AI: How to protect your business in the golden age of fraud and disinformation
• Metrics that matter – How to evaluate identity verification technology
• Digital travel: definition and benefits
• More White Papers

Biometrics Events

• Future Identity Customer London, UK: May 21, 2024
• 2024 SIA GovSummit and Women in Biometrics Awards Washington, D.C.: May 21 - May 22, 2024
• ID4Africa 2024: The Annual General Meeting Cape Town, South Africa: May 21 - May 24, 2024
• EAB Lunch Talk: Implications of the forthcoming forensic sciences standard ISO/IEC 21043 for forensic biometrics Online: May 21, 2024
• Biometrics Institute Asia-Pacific Conference Sydney, Australia: May 22 - May 23, 2024
• More Biometrics Events

SIA Reveals 2024 Winners of SIA Women in Biometrics Awards

The Security Industry Association has named four recipients of the 2024 SIA Women in Biometrics Awards. The 2024 awardees, who will be honored at SIA GovSummit, SIA’s annual government security conference, are:

Leire Arbona Puértolas, Legal and Compliance Director, Veridas

Puértolas has established herself as a leading expert in biometric identity verification and data protection laws since joining Veridas in 2017. She is instrumental in guiding the company’s compliance with international regulations, contributing to Veridas becoming a model for compliance and innovation in digital identity services, influencing industry standards and practices worldwide. Arbona Puértolas’ efforts extend beyond corporate compliance; she actively contributes to the legislative process in various countries, advising on laws that shape the future of digital identity verification. She has played a crucial role in legislative developments in Spain and the European Union, and her expertise is frequently sought in international forums on privacy, biometrics and identity management.

Melissa Conley, Executive Director, Capability Management and Innovation, Requirements and Capabilities Analysis, Transportation Security Administration (TSA)

Conley is the TSA executive director for capability management and innovation (CM&I) for the Office of Requirements and Capabilities Analysis (RCA). She is a member of the Senior Executive Service. Conley provides enterprise oversight for all TSA operational capabilities throughout their life cycles to ensure proper resourcing and alignment to U.S. Department of Homeland Security (DHS) and TSA strategies and priorities. Additionally, Conley coordinates innovation efforts with stakeholders, including current and emerging technology demonstrations intended to enhance operational capabilities and develop efficiencies.

Heather R. Haller, Acting Unit Chief, Quality & Analysis Support Unit, Biometric Services Section, Criminal Justice Information Services (CJIS) Division, Federal Bureau of Investigation (FBI)

A 27-year FBI employee, Heather Haller serves as a supervisory management and program analyst in the Investigative Services Support Unit (ISSU). Haller started her FBI career at the CJIS Division in an entry-level position as a fingerprint examiner and over the years has supported the FBI in a variety of administrative, analytical and research and development positions. Over the last 14 years, she has been primarily focused on facial recognition and has become a subject matter expert on the CJIS Division’s facial recognition program. As an analyst, Haller was instrumental in laying the groundwork to establish the CJIS Division’s Facial Analysis, Comparison and Evaluation (FACE) Operations Services and participated in the development of the Next Generation Identification (NGI) Interstate Photo System (IPS).

Caitlin Kneapler, Ph.D., Technical Integrator, Senior Biometrics Technologist, Office of Biometric Identity Management (OBIM), U.S. Department of Homeland Security (DHS)

Kneapler is a highly skilled and detail-oriented research scientist with over a decade of diverse experience in various fields, including laboratory research, instruction, nonprofit research and analysis, and government agency work. Her expertise spans a wide range of technical and scientific disciplines, with a particular focus on support of biometric and identity management. She serves as a technical interdisciplinary chemist and technical integrator at the OBIM Futures Identity, a position she has held since August 2020. In this role, she is responsible for driving advancements in biometrics, identity and information technology.

“Now in its tenth year, the SIA Women in Biometrics Awards are a globally recognized initiative to honor the outstanding leaders who are fostering change and innovation in biometric identity and security, and this year’s awardees have an impressive array of achievements in biometric technology, policy and compliance, capability management and operations, research, science and analysis across government and industry,” said SIA CEO Don Erickson. “SIA commends Leire Arbona Puértolas, Melissa Conley, Heather Haller and Caitlin Kneapler for their impressive accomplishments in the biometrics field across and thanks them for their important work to advance our industry and make the world safer.”

The 2024 SIA Women in Biometrics Awards honorees were selected from a large pool of accomplished candidates. The awards program will be presented May 21 at a special ceremony during SIA GovSummit in Washington, D.C.

Share This Story

Restricted Content

You must have JavaScript enabled to enjoy a limited number of articles over the next 30 days.

Related Articles

Winners of SIA’s 2020 Women in Biometrics Awards Announced

SIA Announces 2021 Women in Biometrics Award Winners

SIA Opens Nominations for the 2022 Women in Biometrics Awards

Related products.

Biometrics, Crime and Security

Physical Layer Security in Wireless Communications

Intelligence in An Insecure World, 3rd Edition

Get our new emagazine delivered to your inbox every month., stay in the know on the latest security marketplace trends..

Copyright ©2024. All Rights Reserved BNP Media.

Design, CMS, Hosting & Web Development :: ePublishing