Paper 2014/969

Privacy-Preserving Face Recognition with Outsourced Computation

Can Xiang and Chunming Tang

Abstract

Face recognition is one of the most important biometrics pattern recognitions, which has been widely applied in a variety of enterprise, civilian and law enforcement. The privacy of biometrics data raises important concerns, in particular if computations over biometric data is performed at untrusted servers. In previous work of privacy-preserving face recognition, in order to protect individuals' privacy, face recognition is performed over encrypted face images. However, these results increase the computation cost of the client and the face database owners, which may enable face recognition cannot be efficiently executed. Consequently, it would be desirable to reduce computation over sensitive biometric data in such environments. Currently, no secure techniques for outsourcing face biometric recognition is readily available. In this paper, we propose a privacy-preserving face recognition protocol with outsourced computation for the first time, which efficiently protects individuals' privacy. Our protocol substantially improves the previous works in terms of the online computation cost by outsourcing large computation task to a cloud server who has large computing power. In particular, the overall online computation cost of the client and the database owner in our protocol is at most 1/2 of the corresponding protocol in the state of the art algorithms. In addition, the client requires the decryption operations with only $O(1)$ independent of $M$, where $M$ is the size of the face database. Furthermore, the client can verify the correction of the recognition result.