A replay attack detection scheme based on perceptual image hashing

Abstract Perceptual image hashing has mainly been used in the literature to authenticate images or to identify similar contents for image copy detection. In this work, we investigate the introduction of perceptual image hashing to protect biometric systems against digital replay attacks, and we prop...
Ausführliche Beschreibung

Abstract Perceptual image hashing has mainly been used in the literature to authenticate images or to identify similar contents for image copy detection. In this work, we investigate the introduction of perceptual image hashing to protect biometric systems against digital replay attacks, and we propose a novel Replay Attack Detection (RAD) scheme that guarantees the authenticity of the received biometric image on the one hand, and identifies if it has been resubmitted on the other hand. The proposed scheme has the advantages of operating in an uncontrolled environment and does not need the cooperation of end users. It relies on the Challenge/Response principle and combines perceptual image hashing and data hiding techniques. Additionally, a new perceptual image hashing system that performs both content authentication and identification is proposed for fingerprint images. It is based on a shift-invariant calibration signal technique that uses the Discrete Cosine Transform (DCT) and Discrete Sine Transform (DST) of the Differential Block Luminance Mean (DBLM) features to compute the hash sequence. The performance of the proposed system has been evaluated through intensive experiments with real fingerprint images, and the obtained results have shown that the proposed system can provide high performance in terms of authentication and identification. Furthermore, the proposed system outperformed related state-of-the-art image hashing techniques. Ausführliche Beschreibung