Biometric Data Protection

Biometric recognition systems are clearly vulnerable to a variety of software- and model-based attacks that can compromise different stages of the authentication process. Despite their severity, these threats are often neglected and ignored, raising critical questions: have researchers and industry stakeholders deliberately downplayed these attack risks, or is there a lack of awareness regarding the true extent of these threats? Given the serious implications of these vulnerabilities, the scientific community has proposed several countermeasures over the past two decades. Nevertheless, many persistent and sophisticated threats remain unresolved. In this work, we present a new security framework based on chaos theory and cryptographic techniques to protect confidential biometric data. After converting the biometric data into binary digital images, we apply lightweight encryption by applying bitwise Pseudo-Random binary permutation and genetic crossover to return to grayscale, to render the biometric data unintelligible and resistant to all known attacks. Extensive simulations conducted on a large and diverse fingerprint dataset have demonstrated the strong potential of this approach.