A security enhancement scheme for image perceptual hashing

Perceptual hashing as a new technique for the multimedia content security have been widely discussed. However, the robustness requirement of perceptual hashing brings big challenges to its security. In this paper, we propose a secure enhancement scheme for image perceptual hashing based on error correction coding. By employing error correcting code, images with similar perceptual contents can be mapped into the same hash sequence. Consequently, we no longer store original hash sequence in the database of an image authentication system but the derived redundancy-check bits and cipher of perceptual hash. Our scheme can effectively resist the adversary from finding the relationship between the input image and its hash sequence. Therefore, the chosen-plaintext attack can be avoided. Although our scheme may sacrifice some robustness of an image perceptual hashing algorithm, both theoretical analysis and experimental results illuminate that the robustness loss is still acceptable.