Locating payload embedded by group-parity steganography

Payload location has only been studied under simple least significant bit steganography. In this paper, we present an approach to locate payload embedded by a more sophisticated group-parity scheme in which the payload is determined by the parities of disjoint groups of pixels of equal size. The corresponding payload location problem is to identify the pixel groups that determine the payload. We represent the pixels as nodes in a complete graph. Our approach is to partition the graph into nodes that identify the payload by examining modified pixels in a collection of stego images, where each image embeds the payload using the same pixel groups. In the ideal case when the cover images are available, we show that the expected number of stego images needed to partition the graph into the correct pixel groups is approximately eight times the square of the group size times the natural logarithm of the number of pixels. In more practical settings when the cover images are not available, we use cover estimators to locate modified pixels and present a heuristic to partition the graph based on the obtained information. Experimental results are provided to demonstrate the effectiveness of our approach in locating payload embedded by group-parity steganography.