A Variable Neighborhood Search-Based Method with Learning for Image Steganography

Abstract

Image steganography is a security technique that used to hide secret information such as text or image in another cover image. The cover image, including the secret information, seems to be unchanged, and the hidden information can only be recovered by using a particular decoding technique. This paper proposes a variable neighborhood search (VNS)-based method for image steganography. The proposed VNS is combined with the least significant bits method (LSB) and enhanced with a learning process. LSB is the process of adjusting the lower bits of the pixels of the cover image. The least significant bit which is the eighth bit of some or all bytes inside the cover image is replaced by bits of the secret information. We improve LSB by combining it with VNS. The VNS method is a local search meta-heuristic working on a set of different neighborhoods. The basic idea is a systematic change of a certain number of neighborhoods combined with a local search. The objective is to explore the search space efficiently in order to locate the appropriate positions in the cover image where inserting the secret information. Further, a learning process is added to VNS in order to enhance the performance. The proposed methods are evaluated on some series of images. The numerical results are exciting and demonstrate the benefits of the new techniques for image steganography.