Coverless Image Steganography Using Histograms of Oriented Gradients-Based Hashing Algorithm

Abstract

The traditional image steganography approaches usually need a cover image so that the secret information can be imperceptibly embedded into it for secret communication. However, by utilizing the embedding traces left in the cover image, the existence of secret information could be successfully determined by steganalysis methods. To resist the existing steganalysis methods, a coverless image steganography approach is proposed using histograms of oriented gradients (HOGs)-based hashing algorithm. More specially, instead of designating a cover image for secret information embedding, the original images whose hash sequences equal to the secret information are directly selected from a large-scale database, which can be used as the stego-images for secret communication. In this approach, each image's hash sequences are obtained from its non-overlapping blocks by using the HOGs-based hashing algorithm. Both the theoretical analysis and experiments demonstrate that our coverless image steganography approach has good resistance to the existing steganalysis methods, and has good security and strong robustness to the common attacks such as resampling, brightness and contrast change, smoothing as well as addition of Gaussian noise.