Study On Algorithm Of Steganalysis For Still Images

There are many algorithms available to detect the stego images, but much fewer algorithms owe enough credibility under the condition of low-rate-embedding. So, detection accuracy need to be enhanced. Contrast to methods based on spatial domain, methods which are apply to detect hidden message in frequency domain need to be improved and much more sensitive algorithms need to be researched. Besides replacement of LSB, Quantize Index Modulation is another popular method, how to uncover the inherent change due to the action of quantization needs further research. narrow steganalysis refers only to detection of hidden message , but the generalized steganalysis includes extraction of message, so steganography keywords need to be attacked.The Ezstego algorithm based on the replacement of LSB has the characterized color pairs which can be used for detection. The algorithm proposed in this paper for detection the message embedded in the DCT coefficients are based on the prediction and proved effective, even for the advanced algorithms such as F5. There are always round errors due to quantization, which destroys the continuity of coefficient values, So the error distribution function will convert from uniformity distribution to Gaussian distribution. This change uncovers the quantization inherent change for the sake of embedding hidden message. Through building the model of Pseudo Random Number Generator(PRNG) as a directed graph, encryption is also studied in this paper, a stego key algorithm is proposed for the steganograph in the spatial domain.