Abstract:
Biometric recognition systems provide a reliable and automatic way for users to authenticate themselves since they do not need to remember a password or carry a token. However, since biometric data are considered personal and sensitive data, privacy issues arise. Hence, secure biometric systems providing template protection schemes are required. The problem approached in this thesis is to search for an efficient way to use homomorphic encryption for biometric template protection. Employing this encryption, biometric templates are stored and compared in the encrypted domain, ensuring permanent protection. To solve the problem of the computational overload, the most consistent bits of the iris code are sorted to the beginning of the template. Hence, the presented improvements for verification and identification split the iris codes into blocks and are able to make a decision without needing to compare all blocks. Therefore, the execution time is reduced while keeping a high accuracy. The combination of the homomorphic encryption NTRU scheme with specific enhancements yields an usable system in terms of biometric performance and efficiency. Consequently, the privacy issues can be tackled with and security granted.