A hybrid BTP approach with filtered BCH codes for improved performance and security

Authentication of users through their biometric identity has continuously shown ever-increasing importance in our daily lives. However, since biometric data is permanently associated with the users, any loss of this data would be disastrous because it is irrevocable and irreplaceable, unlike traditional passwords and tokens. Therefore, cancelable transformation and bio-cryptosystem are the two main methods used to solve the security and privacy issues associated with biometric authentication systems while preserving accuracy. In this paper, we combined both techniques for biometric template protection (BTP), using the merits and advantages of each approach to compensate for the weaknesses of the other. Specifically, fixed-length vector features generated through kernelized learning are canceled using index-of-max hashing. As a result, a Cancelable Transformation is achieved. The acquired canceled vectorial pairs are then fed into a Bio-Cryptosystem based on fuzzy symmetric encryption and filtered BCH (fBCH) codes for encryption-decryption key pairs and key encoding-decoding, respectively. Hence, double-layered protection coupled with efficient error correction is achieved via symmetric encryption and hashing of the projected vectorial features. The proposed method ensures the creation of parity bits from the adjusted fBCH codewords to fit the length of the canceled vectors and correct the errors associated with the codewords. This allows us to retrieve the precise secret information with a high probability for a genuine user while returning a null or relatively less probability for an imposter user. Extensive experiments on eight FVC2002, FVC2004, and FVC2006 fingerprint datasets are carried out to validate the scheme's performance. Finally, the revocability, unlinkability, and security analysis are experimentally validated.