Robust Image Hashing in Encrypted Domain

In recent decades, a large number of robust image hashing methods have been designed for image authentication. However, existing hashing methods generate hash sequences from the plaintext images. To protect the privacy of image content, we propose a robust image hashing method for encrypted images in this article. In our method, the hash sequence of an image can be extracted from its encrypted version without decryption, and the obtained hash sequence is consistent with that extracted from the corresponding plaintext image. To achieve this, the plaintext image is encrypted using Paillier cryptosystem. Then, according to the homomorphism, hash sequence can be extracted using Walsh-Hadamard transform (WHT) and histogram statistic. For application, a general framework for image authentication in encrypted domain is also presented, including three main entities, i.e., image owner, cloud server and authentication user. An image can be efficiently authenticated by the cloud server without disclosing the image content of the image owner. Compared with typical hashing methods that designed for plaintext domain, our method achieves satisfactory privacy-preserving performance, and also has better performances of image authentication, i.e., robustness and discrimination.