Reliable Detection Technology For Typical Steganographic Algorithm Of Jpeg Images

Steganography is a convert communication technology that hides information in the redundant of multimedia files. It is a new method for secret communication, and owns wide application foreground. Whereas, steganography is two-sided, and it may be used by illegal groups such as QI Qaeda and Falun Gong to endanger the national security and social stabilization. Therefore, steganalysis, that is how to decide whether the detected object contains messages or estimate the length of secret messages, has very important theory worth and practical significance.This dissertation takes JPEG images as covers, and focuses on the researching of embedding message estimation for some steganographic methods, United-Judgment of multiple detection methods. The main works are as follows:For the random additive steganography in DCT domain, considering the alteration of the coefficients as additive noise, and using the variance of differences between neighboring pixels on the block border to describe the border artifacts, the relationship between the border artifacts and the ratio of altered coefficients is derived, which gives support to the design of the steganalytic algorithms for the random additive steganography.For the detection of random additive steganography in DCT domain, an embedding ratio estimation equation is established based on the relationship of the border artifacts and the ratio of altered coefficients. The expression format of the estimation equation is different for different steganographic methods. Taking the embedding ratio estimation of Jsteg and MB1 algorithms as examples, application of the estimation equation is analyzed. Experimental results show that the estimation equation is effective for the embedding ratio estimation of random additive steganography.For the problem that the block border is modified after MB2 steganography and the existed detection methods based on border artifacts is invalidated, new pixels relativity evaluation method is proposed. The approximate linear relationship between the evaluated value and embedding ratio are derived by experiments, and embedding ratio estimation method is proposed based on that. Experimental results show that the proposed steganalytic method is effective.For different characteristics of multiple detection methods under different conditions, according to the principle of setting weights based on the performance, United-Judgment methods for universal blind detection and specific steganalysis are proposed respectively. Experimental results show that the proposed methods can synthetically utilize multiple methods effectively, and achieve more reliable detection.Lastly, the dissertation is concluded and some other problems to be solved are pointed out.