Abstract. In parity assignment-based steganography for palette images, all colors in a host image are divided into two subsets, and each pixel is used to carry one secret bit. This paper describes an analytic method against the parity assignment-based steganographic techniques. By finding the rule of color modifications, a steganalyst can attempt to recover the original histogram in a way that is a reverse of data embedding. Because of the abnormal colors in the original image, an excessive operation will cause some negative values in the recovered histogram. This provides a clue for revealing the presence of secret message and estimating the length of embedded bit sequence.