Customisation of Paillier homomorphic encryption for efficient binary biometric feature vector matching

Computing the Hamming weight between binary biometric feature vectors in a homomorphic encryption domain can be rather inefficient due to the required bit-wise encryption. A biometric matching technique more efficient than the Goldwasser-Mlcali approach is proposed based on exploiting Paillier's capability of encrypting messages larger than one bit at a time. The efficiency is documented in an iris identification context.