A Novel Quantity Based on Clipping Statistics for Jsteg Steganalysis

Steganalysis is the art of detecting covert communication and has become an important research field, in which some of steganographic tools provide the function of embedding secret messages into JPEG images. To detect JPEG stego-images, we will identify a novel statistical quantity in this paper. In general, the JPEG decompression process performs color clipping in lots of pixels, and thus, message embedding process not only degrades the image fidelity but also increases the number of color clipping. When a JPEG image is intercepted on the Internet, the number of color clipping is first computed in the decompression process. Then we embed binary random data into the LSB of non-zero and non-one quantized DCT coefficients for simulating the Jsteg embedding process. Finally, the clipping statistics are evaluated for constructing our steganalytic model. The advantages of the proposed statistical quantity are the simplicity and the ease of implementation. The experimental results show that our clipping statistics are effective to recognize the Jsteg stego-images.