Using permutations to hide information

We describe a new image steganography method that uses permutations of selected pixels to hide information. The basic idea is that a subset of pixels is selected by a keyed random number generator. The values of these pixels are altered by as little as possible but such that the sorted ordering of these values corresponds to a specific permutation. N pixels can be ordered in N ways, corresponding to log2 N bits of hidden information. We explore two different methods of selecting the pixels, one that selects them in a scattered fashion and one that clusters them in small neighborhoods. We also examine the tradeoffs associated with large versus small subsets of pixels.