Threat Taxonomy for Biometric Systems

Authors Dr Ugur Ilker Atmaca, Research Associate Dr Mark Hooper, Technical Development Manager Professor Carsten Maple, Principal Investigator, and Professor of Cyber Systems Engineering with Institute partner University of Warwick (WMG) Acknowledgments This work is supported, in whole or in part, by the Gates Foundation [INV-057591]. Under the grant conditions of the Foundation, a Creative Commons Attribution 4.0 Generic License has already been assigned to the Author Accepted Manuscript. The Alan Turing Institute British Library 96 Euston Road London NW1 2DB Contents 1 Introduction 2 2 Preliminaries 3 2.1 BiometricTraits............................................. 3 2.1.1 PhysiologicalTraits....................................... 3 2.1.2 BehavioralTraits ........................................ 3 2.2 BiometricSystems ........................................... 3 2.2.1 Subsystems ........................................... 4 2.2.2 Processes ............................................ 5 3 Timeline of the Major Incidents 6 4 Threat Taxonomy 9 4.1 SecurityThreats ............................................ 9 4.1.1 Sensor-LevelThreats...................................... 9 4.1.2 Raw Biometric Data-Level Threats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 4.1.3 SupplyChainThreats ..................................... 9 4.1.4 EnvironmentalDisruption................................... 10 4.1.5 Biometric Template-Level Threats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 4.1.6 Communication and Transmission Threats . . . . . . . . . . . . . . . . . . . . . . . . . . 10 4.1.7 AlgorithmicThreats ...................................... 10 4.1.8 OperationalThreats ...................................... 11 4.1.9 InsiderThreats ......................................... 11 4.2 PrivacyThreats............................................. 11 4.2.1 LinkingThreats......................................... 11 4.2.2 IdentificationThreats ..................................... 12 4.2.3 Non-RepudiationThreats ................................... 12 4.2.4 DetectionThreats........................................ 12 4.2.5 DataDisclosureThreats .................................... 12 4.2.6 UnawarenessThreats...................................... 12 4.2.7 Non-ComplianceThreats.................................... 13 4.3 FairnessThreats............................................. 13 4.3.1 BiasinRecognition....................................... 13 4.3.2 UnequalErrorRates ...................................... 13 4.3.3 AccessibilityExclusion..................................... 14 5 Discussion and Summary 14 Page 1 of 16 Threat Taxonomy for Biometric Systems Ugur Ilker Atmaca1, Mark Hooper1, and Carsten Maple1 1The Alan Turing Institute Abstract Biometric traits such as fingerprints, facial features, irises, and voices are unique and advantageous for identifying systems; yet, their permanent nature makes them susceptible to irreversible risks related to security, privacy, and fairness. As the dependence on biometric systems for authentication, surveillance, and identification increases, these challenges are also growing. Biometric systems have faced evolving threats over the past two decades, exposing vulnerabilities and motivating developments in security measures. Early incidents, such as spoofing with "gummy fingers" in 2002 and severe exploits like severed finger cases in 2005, exposed critical weaknesses. More recent threats include DeepFake attacks for biometric fabrication, large-scale breaches, and AI-driven spoofing, prompting innovations like multi-factor authentication and liveness detection. The increase in the access and availability of the AI-based tools caused significant changes in the threat landscape of biometric systems by raising further concerns about privacy, fairness, and ethics, especially with large-scale surveillance systems. Moreover, as quantum algorithms are developed, it becomes more feasible to break existing encryption mechanisms. Thus, there is an emerging requirement for the development of post-quantum cryptographic solutions. This study provides a taxonomy of threats to biometric data and systems, focusing on security breaches, privacy violations, and fairness concerns. It aims to provide a comprehensive list of threats for use in the development of more secure, privacy-preserving, and equitable biometric technologies. 1 Introduction Biometric technologies are essential for modern security and surveillance frameworks for identity verification across a variety of domains, including border control, healthcare, banking, mobile devices, and public safety [1]. While these technologies offer significant advantages, they also present specific vulnerabilities. The evolution of biometric systems and the relevant technologies has shaped its threat landscape as well. While the early demonstration of spoofing used “gummy fingers", recent threats include large-scale data breaches exposing millions of biometric records and deepfake attacks for manipulating facial recognition systems [2, 3]. Threats to biometric systems may compromise security, privacy, and fairness, which may present an interrelationship between their objectives [4, 5]. Security threats, such as unauthorised access or breaches, can compromise the confidentiality, availability, or integrity of sensitive biometric data, exposing individuals to identity theft or system tampering. Privacy threats could compromise "personal information" collected and utilised in the system, which is the information that identifies, relates to, describes, is reasonably capable of being associated with, or could reasonably be linked, directly or indirectly, with a particular consumer or household. Similarly, threats may jeopardise the fairness of the system by embedding bias into the biometric algorithms or providing inequitable access to technology for the various groups or individuals among the society. These threats are not isolated but interdependent and may contain synergies and tensions [6]. For example, security measures may conflict with privacy, as biometric authentication systems rely on the collection and processing of sensitive personal data. An honestbut-curious biometric system may use such data for over-surveillance of society or particular individuals. Similarly, inadequate use of privacy-preserving measures may limit the ability to detect and rectify biases, undermining fairness in the biometric algorithm or the data. Thus, addressing these threats requires a holistic approach that enables tailored integration of the security, privacy, and fairness measures to balance these properties. This report aims to explore the evolving threat landscape of biometric systems and provide a taxonomy of the threats classified as security, privacy, and fairness. Deepfake biometric fabrication and malicious use of quantum algorithms have the potential to significantly change the threat landscape. Moreover, it Page 2 of 16 provides a timeline of the most significant security incidents that have occurred [7, 8, 9]. The scope of this report encompasses an in-depth analysis of biometric systems, their subsystems, and the processes involved in identity verification, based on academic literature, industry reports, and regulatory requirements. The key contributions of this report is as follows: • A review of biometric systems for threat analysis, highlighting entry points and impacts on various processes and subsystems. • An evaluation of emerging threats, including a timeline of major security incidents from the early 2000s to the day. • A comprehensive list of threats to assist the researchers and engineers in developing secure, privacy-preserving, and fair biometric systems. 2 Preliminaries This section provides an overview of biometric systems, including distinct physiological and behavioral traits [10], biometric subsystems as outlined in the ISO/IEC 24745 [11], key processes, and applications. 2.1 Biometric Traits Biometric traits are broadly classified into two main categories as physiological and behavioral. Physiological features refer to bodily characteristics such as fingerprints, face patterns, and iris structure, while behavioral traits involve actions such as vocal patterns, typing rhythm, or stride [12]. These characteristics are employed since they are unique to each person and difficult to replicate. For a biometric trait to be useful, it must meet specific criteria [13]: • Universality: All individuals should possess this trait to ensure widespread applicability. • Uniqueness: The trait must vary considerably among individuals to minimise false matches. • Permanence: The trait must be largely consistent throughout time to ensure long-term reliability. • Measurability: The trait must be easily quantifiable and digitisable by sensors or devices. • Performance: The system should achieve high accuracy with low False Acceptance Rate (FAR) and False Rejection Rate (FRR). • Acceptability: Users must be willing to use the system and provide their biometric data. • Circumvention: The system must resist spoofing or fraudulent manipulation. 2.1.1 Physiological Traits Physiological traits involve the physical attributes of an individual. Typical instances include: • Fingerprint Recognition: Captures the ridge and valley patterns on fingertips for identification. • Facial Recognition: Analyses facial structures, such as the distance between eyes and nose shape, for recognition. • Iris and Retina Scanning: Uses the unique patterns in the iris or blood vessels in the retina for secure identification. • Vascular Patterns: Examines vein structures in hands or wrists for identification, particularly in systems requiring high security. 2.1.2 Behavioral Traits Behavioral traits are based on the unique patterns of actions and interactions of individuals. These include: • Voice Recognition: Analyses vocal patterns based on pitch, tone, and rhythm. • Gait Analysis: Studies walking patterns to identify individuals, often captured via video or floor sensors. • Keystroke Dynamics: Tracks typing speed and rhythm to differentiate users. • Signature Dynamics: Captures the way an individual signs, including pen pressure and stroke order. 2.2 Biometric Systems Biometric systems are vital for existing security frameworks since they provide reliable and robust authentication mechanisms based on individuals’ unique physical and behavioural traits. Thus, they often provide enhanced security to mitigate the unauthorised access threats compared to the security frameworks depending on passwords or tokens. The ISO/IEC 24745 standard outlines a reference architecture for the design and implementation of biometric systems, comprising five interrelated subsystems: biometric data capture, signal processing, data storage, matching, and decision subsystems [11]. Each subsystem executes essential functions that collectively facilitate the biometric Page 3 of 16 system’s precise and efficient implementation of authentication, identification, surveillance or deduplication. Figure 1 illustrates the reference architecture, highlighting integration and flow of information between subsystems. 2.2.1 Subsystems Each subsystem is responsible for a particular step in the biometric data lifecycle, including the capture, processing, storage, comparison, and analysis of biometric data. The roles and interactions of subsystems are also needed for understanding the potential threat entry points and what possible paths could be followed by an adversary and what hazards could occur. Biometric Data Capture Subsystem: This subsystem serves as the primary interface between the user and the biometric system. It employs specific sensors to capture biometric traits and transforms them into biometric samples. These sensors vary depending on the modality in use, such as optical and capacitive fingerprint scanners, cameras for facial recognition, and microphones for voice authentication [14]. The quality and accuracy of the captured data are critical, as they directly affect the system’s performance in the following phases. Advanced sensors have been developed to address environmental challenges such as low-light conditions, motion artefacts, and noise [15]. For instance, multispectral fingerprint scanners enable capturing detailed representations of fingerprints for enhancing recognition accuracy for dirty or damaged fingerprint capturing [16]. Similarly, 3D facial recognition systems and depth cameras provide higher robustness against environmental variations in lighting and facial expressions compared to traditional 2D cameras [17]. Furthermore, noise cancellation and anti-spoofing mechanisms are often employed to enhance the reliability of the data capture process [18]. Signal Processing Subsystem: Once raw biometric data is captured, this subsystem extracts biometric features from the captured sample. It processes the raw biometric data into output numbers or labels that can be compared with those extracted from other biometric samples. This involves preprocessing steps including noise reduction, normalisation, and feature extraction [19]. These traits are then converted into a vectorial representation, which functions as the input for the matching process. The efficacy of the signal processing subsystem is critical to the system’s resilience to fluctuations in lighting conditions, sensor quality, or minor alterations in the biometric trait. Recent studies have demonstrated significant advancements in the processing of biometric templates by utilising deep neural networks (DNNs). For example, pre-trained convolutional neural network (CNN) models such as Inception, Xception and NasNetLarge could achieve recognition accuracy higher than 97% for a variety of fingerprint patterns, including whorls [20]. Data Storage Subsystem: The biometric references obtained in the enrollment process are stored in data storage subsystem. Biometric references are usually stored as biometric templates, which are mathematical representations of the collected features, instead of raw data to prevent reverse-engineering of the original biometric traits [21]. These templates are also often preserved in encrypted forms to prevent unauthorised access and ensure compliance with data protection regulations, such as GDPR. For example, template protection using cancelable biometrics is an effective method for securing stored templates. Techniques such as hashing and salting are employed to build resilient systems. Biohashing methods combine a pseudo-random number generator key with fingerprint minutiae features (e.g., ridge endings and bifurcations) to produce a unique and secure binary code called a Fingerhash [22]. Matching Subsystem: This subsystem compares features extracted from a new biometric sample against the stored references. There are two main types of comparison: verification (1:1 matching), in which the system checks to see if the person providing the biometric sample matches the claimed identity, and identification (1:N matching), in which the system uses a database of references to find out who the person is [23]. This subsystem employs techniques such as pattern recognition, machine learning, and deep learning to achieve high accuracy and minimise false positives or negatives. It is particularly important for biometric deduplication which is gaining further significance for maintaining data integrity in applications such as access control, financial transactions, or humanitarian aid distribution, where duplicate entries can lead to a security, operational, or resource allocation issues [24]. It employs methods for accurately identifying duplicates and maintaining privacy through techniques such as secure multiparty computation. Decision Subsystem: The final determination based on the output of the matching process includes evaluating the similarity scores produced during matching against predefined thresholds [25]. The decision subsystem may also integrate additional contexPage 4 of 16 Fig. 1: Reference Architecture of a Biometric System, taken from [11] (BR (Biometric Reference): Stored data used for comparison; IR (Identity Reference): Non-biometric attributes linked to an individual’s identity.) tual factors, such as the location of the authentication attempt, time of access, or user-specific behavioural traits. This multi-model approach helps minimise errors, reduce false positives or negatives, and ensure a secure and context-aware decision-making process. 2.2.2 Processes The above subsystems facilitate three main processes as follows: Enrollment: It is the initial process of registering an individual’s biometric data into the system. During this phase, the biometric data capture subsystem collects the raw data, which is subsequently processed to extract features and generate a mathematical representation. This template is securely stored in the data storage subsystem, forming the reference against which future samples will be compared. The quality of the enrollment process is critical, as any errors or inaccuracies can adversely impact the system’s performance during verification or identification processes. It serves as the foundational process in biometric reduplication, as unique biometric data is captured, processed, and securely stored in the database for future comparisons. Thus, the accuracy of the reduplication process relies on the accuracy of the enrollment data. Verification: In this process, a one-to-one (i.e., 1:1) comparison is executed for matching newly captured biometric data against the user’s previously stored template to confirm their claimed identity. This is typically used for authentication purposes. The verification process is widely employed in applications such as unlocking personal devices, authorising financial transactions, and securing physical access. Biometric authentication improves simplicity and security by replacing conventional techniques such as passwords and PINs. However, it requires additional measures to protect biometric templates from data breaches, as compromised biometric data cannot be readily changed or reissued like passwords. Identification: This involves a one-to-many (i.e., 1:n) comparison, where the captured biometric data is compared against all stored templates in the database to determine the individual’s identity. While the term "identification" is sometimes used interchangeably with "verification," these are distinct processes. Verification involves a one-to-one (1:1) comparison to confirm that an individual matches the credentials they claim (e.g., a national ID card or passport). In contrast, identification determines who someone is Page 5 of 16 without their prior declaration. Identification could be employed in applications like law enforcement or targeted surveillance, where individuals are matched to known profiles. For instance, casinos may use biometric identification to match clients against a database of blacklisted gamblers [26]. However, national ID programs, electoral registration, or border management are primarily rely on verification to authenticate individuals, ensuring that their identity corresponds to previously established records. Advanced surveillance employs technologies such as facial recognition, gait analysis, and behavioural detection to enhance the large-scale identification of individuals in real time. These systems are implemented in public areas to improve security by detecting possible threats, managing crowds, and monitoring individuals of interest. However, its deployment may raise concerns about data security, potential misuse, and ensuring fairness. 3 Timeline of the Major Incidents The biometric systems evolved over the past two decades - were influenced by many notable incidents. Figure 2 presents a timeline highlighting major security-related incidents in biometric systems, with references provided as footnotes. In 2002, researchers demonstrated the vulnerability of fingerprint systems using artificial “gummy fingers", as an early spoofing technique which highlighted the need for improved biometric security mechanisms. In 2005, car thieves severed a man’s finger to bypass a vehicle’s biometric security system, highlighting the extreme lengths criminals could go to exploit biometric systems. While mitigation techniques have been introduced, over time more sophisticated methods are adopted by the threat actors. In 2013, the Chaos Computer Club successfully demonstrated that Apple’s TouchID could be bypassed by using a fake fingerprint. By 2015, the threat landscape had expanded to include large-scale data breaches, such as the leak of 5.6 million fingerprint records belonging to U.S. government employees. Such incidents have provided lessons and motivation for enhancing the security of biometric systems and the reduction of systems reliance on single points of failure. Techniques include adoption of multi-factor authentication and randomised biometric processes. Additionally, liveness detection algorithms have been adopted to counter evolving spoofing techniques that are becoming more advanced and widespread. The emergence of DeepFake technologies provided a novel and advanced tool to conduct spoofing attacks against biometric systems. DeepFakes are able to generate highly realistic synthetic biometric data that can bypass fingerprint, facial recognition, and voice authentication systems. These capabilities significantly increase risks to authentication processes and potentially enable identity theft to a larger scale. Hence, there is an ongoing need for the development of AIbased detection systems in order to preserve the integrity of biometric security. The integration of AI-driven technologies into biometric surveillance systems has raised new concerns about privacy and fairness. For example, China’s deployment of large-scale AI-enhanced surveillance infrastructure demonstrates the growing role of biometrics in large-scale monitoring systems. While these systems can enhance public safety, they can be potentially used for tracking and controlling populations, requiring transparent legal and ethical measures. Furthermore, a biassed algorithmic decision-making may have negative impacts, particularly for marginalised groups and minorities. For example, some facial recognition algorithms have higher error rates when identifying women and individuals with darker skin tones. This bias has led to cases of wrongful arrests, such as the incident of an African American man who was misidentified by the facial recognition system as a suspect and wrongfully arrested by the Detroit Police Department in the United States [27]. Future biometric threats could further use the convergence of advanced technologies. AI-generated disinformation campaigns, like those influencing elections, may be on the increase. Similarly, deepfake attacks targeting sectors, such as finance suggest that threat actors will continue the use of AI-driven tools to exploit vulnerabilities. Additionally, the increasing use of facial recognition in public spaces, such as stadiums, presents new challenges in balancing security with privacy in terms of linkability threats. Another significant threat is the emergence of quantum computing, which may make it easier to break the current encryption techniques. For example, Shor’s algorithm has the potential for deciphering sensitive biometric information. Hence, there is an increasing need for the adoption of post-quantum cryptographic solutions in biometric systems. AI-driven countermeasures can also enhance the protection against adversarial machine learning attacks. However, the deployment of such technologies must address key ethical concerns, including privacy protection, data security, and algorithmic bias. Clearview AI recieved a $33 million fine for its unauthorised facial recognition database due to recent regulatory actions. Transparency, fairness, and accountability are vital to establishing public trust and realising the potential benefits of these systems. Page 6 of 16 Fig. 2: Timeline of the Major Incidents to Biometric System 1https://www.spiedigitallibrary.org/conference-proceedings-of-spie/4677/1/Impact-of-artificial-gummy-fingers-on-fingerprintsystems/10.1117/12.462719.full 2http://news.bbc.co.uk/1/hi/world/asia-pacific/4396831.stm 3https://www.ccc.de/en/updates/2007/umsonst-im-supermarkt 4https://www.ccc.de/en/updates/2013/ccc-breaks-apple-touchid 5https://www.theguardian.com/technology/2014/dec/30/hacker-fakes-german-ministers-fingerprints-using-photos-of-her-hands 6https://www.nytimes.com/2015/09/24/world/asia/hackers-took-fingerprints-of-5-6-million-us-workers-government-says.html 7https://www.voanews.com/a/fingerprint-photos/3678345.html 8https://www.ftc.gov/enforcement/refunds/equifax-data-breach-settlement 9https://www.washingtonpost.com/news/worldviews/wp/2018/01/04/a-security-breach-in-india-has-left-a-billion-people-at-risk-ofidentity-theft/ 10https://thewire.in/economy/aadhaar-fraud-uidai 11https://iapp.org/news/a/how-to-interpret-swedens-first-gdpr-fine-on-facial-recognition-in-school 12https://www.crowdstrike.com/en-us/cybersecurity-101/social-engineering/deepfake-attack/ 13https://3dprint.com/271540/d-defcon-fooling-biometric-sensors-using-3d-printed-fake-fingerprints/ 14https://www.wired.com/story/cheap-3d-printer-trick-smartphone-fingerprint-locks/ 15https://www.reuters.com/world/china/china-uses-ai-software-improve-its-surveillance-capabilities-2022-04-08/ 16https://economictimes.indiatimes.com/tech/technology/aadhar-data-leak-personal-data-of-81-5-crore-indians-on-sale-on-dark-webreport/articleshow/104856898.cms?from=mdr 17https://www.cyberpeace.org/resources/blogs/aadhaar-biometric-fraud 18https://www.biometricupdate.com/202312/missing-fingerprint-liveness-detection-opens-the-door-for-aadhaar-spoofing-attacks 19https://www.biometricupdate.com/202312/biometric-verification-system-fails-to-distinguish-identical-twins-in-ghana-local-elections 20https://www.biometricupdate.com/202411/biometrics-coming-to-more-stadiums-with-facial-recognition-tender-in-nsw 21https://cybernews.com/news/bypassing-biometric-facial-recognition/ 22https://arstechnica.com/information-technology/2023/05/hackers-can-brute-force-fingerprint-authentication-of-android-devices/ 23https://www.forbes.com/sites/roberthart/2024/09/03/clearview-ai-controversial-facial-recognition-firm-fined-33-million-for-illegaldatabase/ 24https://reutersinstitute.politics.ox.ac.uk/news/how-ai-generated-disinformation-might-impact-years-elections-and-how-journalistsshould-report 25https://www.theguardian.com/australia-news/2024/nov/19/bunnings-facial-recognition-technology-breach-stores-ntwnfb 26https://www.wsj.com/articles/deepfakes-are-coming-for-the-financial-sector-0c72d1e5 Page 7 of 16 uals of varying skin tones have led to substantial discrepancies in error rates, often disadvantaging minority groups [35, 36]. 4.3.3 Accessibility Exclusion This refers to the exclusion of individuals with physical impairments or non-standard biometric traits. For example, individuals with missing fingers, facial disfigurements, or unique physical conditions may find themselves excluded from systems that rely on standard biometric features, raising concerns about their fairness. Biometric systems often depend on templates derived from usual traits, and they may neglect outliers with non-standard traits. For instance, people with disfigurements on their faces might be misclassified or even rejected outright. People with unique physical traits, such as genetic conditions that change biometric patterns, may also have more problems [37, 38]. 5 Discussion and Summary The significance of biometric systems for security and surveillance frameworks is growing, as they are increasingly utilised for identity verification in various sectors, such as healthcare, banking, border control, and public safety. This report presents a comprehensive taxonomy of threats to biometric systems, emphasising the interconnected challenges of security, privacy, and fairness. For instance, security threats could be used to conduct algorithmic manipulation or template theft, which are fairness and privacy issues, respectively. Thus, it is required to address these threats in a comprehensive way rather than focusing on one of them. Furthermore, the report reviews the evolution of these threats through a timeline of incidents, from early spoofing demonstrations to sophisticated deepfake biometric generation attacks. The threat categories are as follows: – Security Threats: These include physical and digital threats, such as spoofing, sensor tampering, and algorithmic manipulations, as well as emerging threats such as deepfake biometric fabrication and malicious use of quantum algorithms. – Privacy threats: These include concerns about the misuse and unauthorised access of biometric data, leading to issues including linking, identification, nonrepudiation, detection, data disclosure, unawareness, and non-compliance. – Fairness Threats: These include threats causing bias in recognition algorithms, unequal error rates, and accessibility exclusions that disproportionately affect certain demographics. In addition to the identified threats, there are emerging tools and techniques that pose novel ways to attack biometric systems. Deepfake technologies are threatening the security of biometric systems by creating real-like synthetic biometric traits [3]. An adversary may use AI-generated synthetic biometric traits during data capture, creating fake biometric data, such as realistic faces, voices, or fingerprints. Additionally, deepfake technology can facilitate generating manipulated video or audio content [39, 40]. However, a wide range of countermeasures are also integrated into biometric systems to enhance the system’s resistance to spoofing, data breaches, and privacy violations. Liveness detection mechanisms verify that the biometric sample originates from a living individual rather than a spoofed artefact, such as masks, fake fingerprints, or deepfake videos [41]. Another critical challenge is the emerge of the quantum algorithms which can undermine existing cryptographic foundations used in protection of biometric data [42]. Quantum algorithms, such as Shor’s algorithm [43], will be able to break the majority of existing encryption methods once sufficiently advanced quantum computers become available. This capability would enable attackers to decrypt stored biometric templates and compromise data confidentiality, highlighting the urgent need for the development and adoption of quantum-resistant cryptographic solutions for maintaining the confidentiality. These algorithms could decrypt previously secure biometric templates, exposing sensitive personal information, as encrypted data could be collected today and decrypted in the future as the technology advances. In the biometric matching process, compromised templates may result in unauthorised access, identity theft, and a loss of trust in biometric authentication systems. However, there are ongoing efforts to standardise post-quantum cryptography (PQC) methods to accelerate the adoption of quantum-resistant algorithms [44]. Government agencies, including the U.S. Customs and Border Protection, are already integrating PQC to safeguard sensitive data, including biometrics [45]. Additionally, innovative quantumresistant biometric solutions, such as the Irreversibly Transformed Identity Token are being developed [46]. Ensuring secure, private, and fair biometric systems remains a significant challenge, given the multifaceted vulnerabilities in data protection, the potential for misuse or unauthorised access, and the persistent biases that disproportionately impact underrepresented demographic groups. The implementation of robust privacy-preserving techniques and assured fairness are a must to maintain system performance and trustworthiness. Page 14 of 16 References [1] U Sumalatha, K Krishna Prakasha, Srikanth Prabhu, and Vinod C Nayak. A comprehensive review of unimodal and multimodal fingerprint biometric authentication systems: Fusion, attacks, and template protection. IEEE Access, 2024. [2] Tsutomu Matsumoto, Hiroyuki Matsumoto, Koji Yamada, and Satoshi Hoshino. Impact of artificial" gummy" fingers on fingerprint systems. In Optical security and counterfeit deterrence techniques IV, volume 4677, pages 275–289. SPIE, 2002. [3] Sreeraj Ramachandran, Aakash Varma Nadimpalli, and Ajita Rattani. An experimental evaluation on deepfake detection using deep face recognition. In 2021 International Carnahan Conference on Security Technology (ICCST), pages 1–6. IEEE, 2021. [4] Patrizio Campisi. Security and privacy in biometrics: towards a holistic approach. In Security and privacy in biometrics, pages 1–23. Springer, 2013. [5] Anil K Jain, Debayan Deb, and Joshua J Engelsma. Biometrics: Trust, but verify. IEEE Transactions on Biometrics, Behavior, and Identity Science, 4(3):303–323, 2021. [6] Josep Domingo-Ferrer and Alberto Blanco-Justicia. Ethical valuecentric cybersecurity: A methodology based on a value graph. Science and Engineering Ethics, 26(3):1267–1285, 2020. [7] Chris Roberts. Biometric attack vectors and defences. computers & security, 26(1):14–25, 2007. [8] Yogendra Narain Singh and Sanjay Kumar Singh. A taxonomy of biometric system vulnerabilities and defences. International journal of biometrics, 5(2):137–159, 2013. [9] Sani M Abdullahi, Shuifa Sun, Beng Wang, Ning Wei, and Hongxia Wang. Biometric template attacks and recent protection mechanisms: A survey. Information Fusion, 103:102144, 2024. [10] Paulo Henrique Pisani, Abir Mhenni, Romain Giot, Estelle Cherrier, Norman Poh, André Carlos Ponce de Leon Ferreira de Carvalho, Christophe Rosenberger, and Najoua Essoukri Ben Amara. Adaptive biometric systems: Review and perspectives. ACM Computing Surveys (CSUR), 52(5):1–38, 2019. [11] International Organization for Standardization (ISO) and International Electrotechnical Commission (IEC). Information security, cybersecurity and privacy protection — biometric information protection, 2022. British Standard Implementation: BS ISO/IEC 24745:2022. [12] Anil K Jain and Ajay Kumar. Biometric recognition: an overview. Second generation biometrics: The ethical, legal and social context, pages 49–79, 2012. [13] Muhtahir O Oloyede and Gerhard P Hancke. Unimodal and multimodal biometric sensing systems: a review. IEEE access, 4:7532–7555, 2016. [14] Rodrigo de Luis-Garcıa, Carlos Alberola-Lopez, Otman Aghzout, and Juan Ruiz-Alzola. Biometric identification systems. Signal processing, 83(12):2539–2557, 2003. [15] Chongyi Li, Chunle Guo, Linghao Han, Jun Jiang, Ming-Ming Cheng, Jinwei Gu, and Chen Change Loy. Low-light image and video enhancement using deep learning: A survey. IEEE transactions on pattern analysis and machine intelligence, 44(12):9396–9416, 2021. [16] Robert K Rowe, Kristin A Nixon, and Stephen P Corcoran. Multispectral fingerprint biometrics. In Proceedings from the Sixth Annual IEEE SMC Information Assurance Workshop, pages 14–20. IEEE, 2005. [17] Lale Akarun, Berk Goekberk, and Albert Ali Salah. 3d face recognition for biometric applications. In 2005 13th European Signal Processing Conference, pages 1–5. IEEE, 2005. [18] P Venkata Reddy, Ajay Kumar, SMK Rahman, and Tanvir Singh Mundra. A new antispoofing approach for biometric devices. IEEE transactions on biomedical circuits and systems, 2(4):328–337, 2008. [19] Josef Ström Bartunek, Mikael Nilsson, Benny Sallberg, and Ingvar Claesson. Adaptive fingerprint image enhancement with emphasis on preprocessing of data. IEEE transactions on image processing, 22(2):644–656, 2012. [20] Claudia Cordos ,, Laura Mihail˘ a, Paul Faragó, and Sorin Hintea. Ecg signal classification using convolutional neural networks for biometric identification. In 2021 44th international conference on telecommunications and signal processing (TSP), pages 167–170. IEEE, 2021. [21] Arpita Sarkar and Binod K Singh. A review on performance, security and various biometric template protection schemes for biometric authentication systems. Multimedia Tools and Applications, 79(37):27721–27776, 2020. [22] Mamatha Ram and KR Radhika. Biohashing application using fingerprint cancelable features. In 2020 IEEE 15th International Conference on Industrial and Information Systems (ICIIS), pages 214–218. IEEE, 2020. [23] Zhang Rui and Zheng Yan. A survey on biometric authentication: Toward secure and privacy-preserving identification. IEEE access, 7:5994– 6009, 2018. [24] Kasra EdalatNejad, Wouter Lueks, Justinas Sukaitis, Vincent Graf Narbel, Massimo Marelli, and Carmela Troncoso. Janus: Safe biometric deduplication for humanitarian aid distribution. In 2024 IEEE Symposium on Security and Privacy (SP), pages 655–672. IEEE, 2024. [25] Jyoti Malik, Dhiraj Girdhar, Ratna Dahiya, and G Sainarayanan. Reference threshold calculation for biometric authentication. IJ Image, Graphics and Signal Processing, 2:46–53, 2014. [26] IpKin Anthony Wong, Keng Fong Chau, and Heng U Chan. An empirical study on customers’ gambling intention in ai-supported casinos. Journal of Hospitality and Tourism Technology, 14(2):121–136, 2023. [27] The Guardian. Detroit police facial recognition system sparks racial bias controversy. The Guardian, July 2024. Accessed: 2024-12-22. [28] Leonidas Spinoulas, Mohamed E Hussein, David Geissbühler, Joe Mathai, Oswin G Almeida, Guillaume Clivaz, Sébastien Marcel, and Wael Abdalmageed. Multispectral biometrics system framework: Application to presentation attack detection. IEEE Sensors Journal, 21(13):15022–15041, 2021. [29] Craig W Thompson and Dale R Thompson. Identity management. IEEE Internet Computing, 11(3):82–85, 2007. [30] Antonio Robles-González, Javier Parra-Arnau, and Jordi Forné. A linddun-based framework for privacy threat analysis on identification and authentication processes. Computers & Security, 94:101755, 2020. [31] Livinus Obiora Nweke, Mohamed Abomhara, Sule Yildirim Yayilgan, Debora Comparin, Olivier Heurtier, and Calum Bunney. A linddunbased privacy threat modelling for national identification systems. In 2022 IEEE Nigeria 4th International Conference on Disruptive Technologies for Sustainable Development (NIGERCON), pages 1–8. IEEE, 2022. [32] Pawel Drozdowski, Christian Rathgeb, Antitza Dantcheva, Naser Damer, and Christoph Busch. Demographic bias in biometrics: A survey on an emerging challenge. IEEE Transactions on Technology and Society, 1(2):89–103, 2020. [33] Joseph P Robinson, Gennady Livitz, Yann Henon, Can Qin, Yun Fu, and Samson Timoner. Face recognition: too bias, or not too bias? In Proceedings of the ieee/cvf conference on computer vision and pattern recognition workshops, pages 0–1, 2020. [34] David Leslie. Understanding bias in facial recognition technologies: An explainer. The Alan Turing Institute, 2020. [35] Jacqueline G Cavazos, P Jonathon Phillips, Carlos D Castillo, and Alice J O’Toole. Accuracy comparison across face recognition algorithms: Where are we on measuring race bias? IEEE transactions on biometrics, behavior, and identity science, 3(1):101–111, 2020. [36] Ignacio Serna, Alejandro Pena, Aythami Morales, and Julian Fierrez. Insidebias: Measuring bias in deep networks and application to face gender biometrics. In 2020 25th International Conference on Pattern Recognition (ICPR), pages 3720–3727. IEEE, 2021. [37] Philipp Terhörst, Jan Niklas Kolf, Marco Huber, Florian Kirchbuchner, Naser Damer, Aythami Morales Moreno, Julian Fierrez, and Arjan Kuijper. A comprehensive study on face recognition biases beyond demographics. IEEE Transactions on Technology and Society, 3(1):16–30, 2021. Page 15 of 16 [38] Seyma Yucer, Furkan Tektas, Noura Al Moubayed, and Toby Breckon. Racial bias within face recognition: A survey. ACM Computing Surveys, 57(4):1–39, 2024. [39] Hemlata Tak, Jee-weon Jung, Jose Patino, Madhu Kamble, Massimiliano Todisco, and Nicholas Evans. End-to-end spectro-temporal graph attention networks for speaker verification anti-spoofing and speech deepfake detection. arXiv preprint arXiv:2107.12710, 2021. [40] Hemlata Tak, Massimiliano Todisco, Xin Wang, Jee-weon Jung, Junichi Yamagishi, and Nicholas Evans. Automatic speaker verification spoofing and deepfake detection using wav2vec 2.0 and data augmentation. arXiv preprint arXiv:2202.12233, 2022. [41] Emma Lavens, Davy Preuveneers, and Wouter Joosen. Mitigating undesired interactions between liveness detection components in biometric authentication. In Proceedings of the 18th International Conference on Availability, Reliability and Security, pages 1–8, 2023. [42] Gabriela Mogos. Quantum matching algorithm for biometric fingerprints. In Proceedings of the 2024 8th International Conference on Machine Learning and Soft Computing, pages 197–201, 2024. [43] Artur Ekert and Richard Jozsa. Quantum computation and shor’s factoring algorithm. Reviews of Modern Physics, 68(3):733, 1996. [44] Gorjan Alagic, Gorjan Alagic, Daniel Apon, David Cooper, Quynh Dang, Thinh Dang, John Kelsey, Jacob Lichtinger, Yi-Kai Liu, Carl Miller, et al. Status report on the third round of the nist post-quantum cryptography standardization process. 2022. [45] U.S. Customs and Border Protection. Cbp tackles post-quantum cryptography. November 2024. Accessed: 2024-12-02. [46] Ming Jie Lee, Andrew Beng Jin Teoh, Andreas Uhl, Shiuan-Ni Liang, and Zhe Jin. A tokenless cancellable scheme for multimodal biometric systems. Computers & Security, 108:102350, 2021. Page 16 of 16