Feature Extraction Of Digital Stego-image And Information Detection Methods

Steganography uses human visual sensitivity and digital multimedia signals inherent redundancy to embed secret information into a digital image. This technology helps the national security sector and the military departments encrypt information and transmit information. Steganalysis is a technique of detecting the steganography in digital images. Today, it is a new emerging domain of information security in the world. National research institutions, intelligence agencies, military and government attach great importance to it.This paper presents several feature extraction algorithms and combines them with several classification identifiers in order to construct steganography detection models. These models are used to detect encryption information which used least significant bit steganography algorithm. Steganalysis's conditions need to be treated differently:full-reference images and no-reference images. The main task is presented as follows:1. With full reference images, this article will analyze C4, S4S5and H feature sets, extracts and normalizes them. Some modified images added to the train and test samples. The experimental results show much better detective performance than the regression models'results.2. In the condition of no-reference images, this article presents a new algorithm to extract image features:correlation of wavelet frequency domain. This algorithm finds out that the correlation between the sub-frequency domains after wavelet transforms and extracts the characteristic parameters. The experimental results reflect its applied value.3. In the condition of no-reference images, this article applies the pattern recognition methods for subtractive pixel adjacency matrix feature extraction algorithm and constructs steganography detection models. Experimental results are better than the previous experiments'.