Decoding Multimodal Biometrics in Forensic Investigations

By Heena Chauhan, Budding Forensic Expert

Introduction

Biometric identification is no longer limited to a single feature like fingerprints or face characteristic in modern forensic science. Multimodal biometrics combining two or more biometric traits (like face + iris, or fingerprint + vein) have better accuracy and reliability and are more secure against spoofing. That means, for criminal investigations, better confirmation of identity, fewer false positives and more evidence to back up findings [1].

What is Multimodal Biometrics?

Multimodal biometrics refers to systems that combine more than one way of data such as a physiological approach (like iris or fingerprinting) along with a behavioral method (like voice or gait) to verify identity [1]. Multimodal systems rather than unimodal, make substantial cut down the failure due to bad quality or malicious attacks in biometrics [1].

Advances in Forensic Multimodal Biometrics

1. Face + Iris Fusion
   Face recognition and iris patterns are combined in the majority of forensic biometric systems. For example, one study suggested a multimodal system in which facial features are extracted by singular spectrum analysis (SSA) and iris features are extracted using a multi-resolution Log-Gabor filter, and the two types of feature vectors are separately fused at score level and decision level for better reliability.
   The ability to fuse these two feature types (score-level and decision-level fusion) is most useful in law enforcement when only partial or poor-quality biometric data is available [2].
2. Deep Learning–Based Tri-Modal Recognition
   Recently, CNN has been employed to fuse iris, face, finger-vein data. One well-performing system scored the 99.39% with feature-level fusion and 100% with score-level fusion. This technique enhances robustness in practical forensic environments [1].
   Such systems can operate even when one modality is compromised (e.g., fingerprint smudged), because the other traits back them up.
3. Fingerprint + Iris Biometric Systems
   One of the other well known multimodal approach combines fingerprint with iris biometrics by fuzzy logic or weighted sum score fusion. This decision level fusion enhances recognition performances and provides robustness against spoofing or input biometric data loss [3].

Forensic Applications

• Crime Scene Investigation: Multimodal biometrics can ensure identity with more certainty which is required however at the same time two or more traits must match so that false positive rate will be less [2].
• Suspect Identification: For the identification of a suspect by using a biometric sample and collecting other biometrics (e.g., arrestees in custody/suspects whose identity is in question), the potential evidence value of multiple traits can be considerably more significant [1].
• Mass Disaster and Identification: In disaster situations with multiple casualties like earthquakes, biometrics for face, iris or finger prints can be combined to help identify the individual even if one of the modalities is degraded [4].
• Long-Term Observation: In forensic databases, multi-biometric templates produce a more comprehensive identity profile and can thus be particularly useful for long-term surveillance against repeat offenders or correlating criminal activities [3].

Challenges and Limitations

• Complexity of Data Fusion: The data from the varying body parts (face, iris, vein) to be combined is required some sophisticated fusion algorithms. Bad fusion can decrease accuracy instead of increasing it [1].
• Privacy & Ethics: Collecting more than one type of your biometric factor is a bigger privacy threat. Storing and securing this sensitive information requires strict security measures [5].
• Hardware & Cost: It’s costlier and more complicated to introduce new hardware capturing multiple traits (i.e., camera + iris scanning) [4].
• Variability & Quality: Every biometric has it own vulnerabilities (finger smudges, poor lighting, occluded iris) which can affect performance [2].

Future Directions

• AI-Based Fusion Models: Incorporating AI (Deep Learning) techniques for optimal feature-level as well as score-level fusion will significantly enhance system performance and adaptability [1].
• Portable Forensic Biometric Systems: These are mobile biometric (of several types like face, iris, and fingerprint) devices that capture quality, multisensory data directly out in the field [4].
• Biometric Template Protection: Work on cancelable biometrics and secure biometric cryptographic keys guarantees the privacy of fusion templates and its irreversibility [5]
• Behavioral + Physiological Hybrid Systems: Combining behavioral characteristics (such as gait and voice) with physiological ones (iris, fingerprint) for even more powerful forensic recognition systems [3].

References

1. Alay, N., & Al-Baity, H. H. (2020). Deep learning approach for multimodal biometric recognition system based on fusion of iris, face, and finger vein traits. Sensors, 20(19), 5523. https://doi.org/10.3390/s20195523
2. Laboratory, NDT & University of Paris 8. Face–Iris Multimodal Biometric Identification System. Electronics, 9(1), 85. https://doi.org/10.3390/electronics9010085
3. Benaliouche, H., & Touahria, M. (2014). Comparative study of multimodal biometric recognition by fusion of iris and fingerprint. ScientificWorldJournal, 2014, 829369. PMC
4. Kamble, S. N., Gund, V. D., & Kazi, K. S. (2018). Multimodal biometrics authentication system using fusion of fingerprint and iris. International Journal of Trend in Scientific Research and Development, 2(6), 1282–1286. https://www.ijtsrd.com/papers/ijtsrd18861.pdf
5. Jagadeesan, A., & Duraiswamy, K. (2010). Secured cryptographic key generation from multimodal biometrics feature-level fusion of fingerprint and iris. arXiv. https://arxiv.org/abs/1003.3894