Secure Steganography Using Gabor Filter and Neural Networks

Abstract

The main concern of steganography (image hiding) methods is to embed a secret image into a host image in such a way that it causes minimum distortion to the host; to make it possible to extract a version of secret image from the host in such a way that the extracted version of secret image be as similar as possible to its original version (this should be possible even after usual attacks on the host image), and to provide ways of embedding secret images with larger size into a given host image. In this paper we propose a method that covers all above mentioned concerns by suggesting the idea of finding from an image data base, the most suitable host for a given secret image. In our method, the secret and host images are divided into blocks of size 4×4. Each block in secret image is taken as a texture pattern for which using Gabor filter, the most similar block is found among the blocks of host image candidates. Using this similarity criterion and Kohonen neural network, the most suitable host image is selected from an image database. Embedding is done by placing the blocks of secret image on their corresponding blocks in the selected host image. The location addresses of blocks in host that were replaced by blocks of secret image are saved. They are converted to a bit string that is embedded in DCT coefficients of the hybrid image. Our experimental results showed a high level of capacity, robustness and minimum distortion when using standard images as secret and host images.