Huffman-code based retrieval for encrypted JPEG images

Abstract

This paper proposes a retrieval scheme for encrypted JPEG images, based on Huffman code in the JPEG bitstream. Three parties are involved: content owner, cloud server, and authorized user. First, the content owner produces encrypted images by jointly using a stream cipher, permutation cipher, then uploads them to a cloud server. Especially, the conversion between encrypted quantization tables is still valid. With the same secret key, the authorized user submits the encrypted query image to the server. Second, the server extracts Huffman-code histogram from the encrypted image as a feature. Although the Huffman-code histogram is changed during encryption, encrypted images with similar content to the query image are returned to the user after feature comparison. Finally, through decryption and hash verification, the user can obtain authenticated plaintext images. Experimental results show that the proposed scheme ensures confidentiality, integrity and format compatibility, while image retrieval of different quality factors is still effective.