LAUSR

Privacy and security are main concerns in face authentication by cloud or semi-honest server. To deal with this problem, a face authentication method based on correlation filters in encrypted domain is presented in this paper. This method includes enrollment and authentication steps using homomorphic cryptosystem where public key and correlation filter are shared between client and server. Such system obviates the need for trusting the server and raises a privacy issue by performing all computation in the encrypted domain without needing any decryption, even during face authentication process. In the proposed method, correlation operation, peak to sidelobe ratio (PSR) measurement, and threshold comparison are performed by server in the encrypted domain. We show that the proposed method is secure against typical encryption attacks. Experimental results on LFW, Yale-B, and FERET face databases demonstrate that, the recognition rate in the encrypted domain is almost the same as plain domain.