| Steganography has an important significance embranchment in military and national security. Steganalysis of the current research mainly focuses on detecting the existence of hidden information. But, how to extract hiding information, is playing a decisive role of both catching information in information wars and defeating network crime in computer forensics.Extracting attack, which hardly any literature discuss about, is an inherently difficult problem. The present paper did groping research on the extracting attack to image steganography under stego-only assumption, and got the following results:1. An extracting method for plaintext sequential LSB-replacement and LSB-matching steganography. By using the statistical properties of the top bits of plaintext, a detecting and extracting attack to plaintext sequential LSB steganography is proposed. Both theoretic analysis and experimental results show that the proposed method is very effective. Even in condition of only200bytes English plaintext embedded in an image, the proposed method can detect and extract the plaintext successfully.2. Extracting method for LSB-replacement steganography based on Arnold transformation. First, we discusses the periodicity of Arnold transformation. Then, analyzing the weakness of the steganography algorithm. Finally, we present a key recovering method. Both theoretic analysis and experimental results prove the computational complexity of the proposed method is lower than Fridrich and can recover the stego key successfully.3. Locating steganographic payload for JPEG decompressed images based on hypothesis testing. The original cover was estimated by recompressing the stego image, and then the probability of successfully recovering the original cover was evaluated. The payload location problem was then translated into a hypothesis testing problem, and a payload location method for LSB-replacement and LSB-matching steganography was proposed. Experiments showed that the proposed method is far better than existing algorithms in two ways. First, given the number of stego images, the number of perfectly located load-carrying pixels is close to the theoretical upper bound. Second, the number of stego images needed to locate perfectly all load-carrying pixels is close to the theoretical lower bound.4. Stego key searching for JPEG decompressed images. First, recovering stego key is translated into a problem of cryptanalysis to recover the seed of a Pseudo-Random Number Generator (PRNG) by its sample sequence which contains error bits at the probability of Pe. Then, discuss the two following questions:1) what condition of the error rate probability Pe and the embedded rate r is required to ensure the success of the collision attack;2) what is the minimum sample size with which we can achieve the collision attack. Finally, by using the probability of successfully recovering the original cover of part3, we present a key recovering method based on collision attack. The results of the experiment on steganographic software "Hide and Seek4.1" and its varieties prove the proposed method is effective for both LSB-replacement and LSB-matching and can recover the stego key successfully.5. In addition, based on the result of part3and part4, we also studied the reversible data hiding method based on JPEG decompressed Images. First, this paper proposes the concept of recoverable block, and discusses the relation between the number of recoverable block and image quality factor. Second, through analyzing the properties of DCT coefficients and±k embedding, we get the theoretical probability of successfully embedded, and then obtain the optimum number of embedding pixel and embedding capacity calculation formula. Finally, a new adaptive reversible data hiding algorithm in JPEG decompressed images is proposed. This algorithm outperforms existing methods in terms of high embedding capacity and low distortion. Both theoretic analysis and experimental results show the effectiveness of the proposed algorithm. |
PDF Full Text Request