Crypto-space image steganography based on blind flipping

Abstract

Reversible data hiding (RDH) for digital images is a widely studied steganography that ensures the carrier image can be restored without distortion. Due to the popularity of Internet applications, a lot of research reports focus on the RDH for encrypted images (RDHEI) recently. For a fully encrypted image, data hiding by direct modification of carrier units usually suffers from the problem of error restoration. We propose two mechanisms to solve this problem. The first mechanism is to disperse error units with Arnold transform. The second mechanism is to correct possible errors with Hamming coding. By incorporating both mechanisms, the error restoration rate can be greatly reduced. Besides, the periodicity of Arnold transform is also investigated. By leveraging a full search program and the frequency analysis, the maximum periods together with their parameters for some practical matrix sizes are found. This information is beneficial to get desired matrix state with Arnold Transform. Ninety-eight grayscale images from USC-SIPI and BOSSbase v1.01 image datasets were tested. Due to the reduction of error restoration rate, the proposed mechanisms successfully raise the data payload from 0.1177 to 0.172 bits per pixel.