FaceIDP : Face Identification Differential Privacy via Dictionary Learning Neural Networks

In big-data era, large amount of facial images could be used to breach the face identification system, which demands effective Face IDentification Differential Privacy (FaceIDP) of the facial images for widespread adoption of the face identification technique. In this paper, to our best knowledge, we take the first step to systematically study an effective important FaceIDP approach via the help of Dictionary Learning (DL) for secure releasing of facial images. First, a Dictionary Learning neural Network (DLNet) has been developed and trained with the facial images database, to learn the common dictionary basis of the facial image database. Then, the coding coefficients of the facial images are obtained. After that, the sanitizing noise is added to the coding coefficients, which obfuscates the facial feature vector that is used to identify a user's identification. We have also proved that the FaceIDP is ϵ-differentially private. More importantly, optimal noise scale parameters have been obtained via the Lagrange Multiplier (LM) method to achieve better data utility for a given privacy budget ϵ. Finally, substantial experiments have been conducted to validate the efficiency of the FaceIDP with two real-life facial image databases.