Blind image steganalysis using features extraction and machine learning

The main theme of this work is the steganalysis of image steganography techniques by classification and feature extraction methods. Our contribution is to improve a steganalysis algorithm and implement it to extract the characteristics of the Cover and Stego images. Thus the use of a modern Machine Learning "Ensemble Classifier" is presented here as a powerful development tool that allows the rapid construction of steganographic detectors with significantly improved detection accuracy in a wide range of integration methods. The power of the proposed framework is demonstrated by a steganographic method nsF5 which masks one or more messages in JPEG images.