Study On LSB Steganalysis Algorithm

With the rapid development of the computer and the Internet, the digital multimedia technology has been entered into a new era. Steganography and steganalysis are drawing more and more attentions from governments, military and research institutes all over the world. Steganography can be used to transmit secret files via Internet. But it can also be easily used by malicious groups for evil purpose, such as spreading the illegal information. Therefore, the research on steganalysis is very important to national communication security.LSB steganography is widely used in the Internet, due to its simple implementation and large capitation of hiding information. The thesis of this paper aims to the steganalysis technique of images.In order to detect the LSB replacement, a fast and effective steganalysis method is proposed based on the correlation of neighbor image pixels by using the difference between the oddity and evenness of the adjacent pixels in original images and stego images. The algorithm needs only to count the number of the adjacent pixel pairs in which the odd one is greater than the even, and the number of the adjacent pixel pairs in which the even one is great than the odd, respectively, when the difference of adjacent pixel is positive odd in both the image and its flipping version. The existence of a secret message in the carrier image is determined by SVM (Support Vector Machine) according to the ratio of it. The amount of hidden messages is estimated by SVR (Support Vector Regression). The experimental results show that detection performance is satisfied, especially with the speed of implementation. The new algorithm can be implemented conveniently and has lower computation complexity than SPA (Sample Pairs Analysis).For the steganography method whose embedding message does not be deposed into random form, a LSB steganalysis method is presented based on the complexity of least significant bit. The complexity of LSB obeys normal distribution. Because the LSB of original image is like as random noise, and the probability of bit value is 1/2, which meets Central Limit Theorem. The algorithm exploits the differences between the complexity of the least significant bit of original images and stego images. The new algorithm provides a new direction for LSB steganalysis, and the experimental results show that the detection performance is effective and reliable.