A ±1-Based Steganography by Minimizing the Distortion of First Order Statistics

It is well-known that least significant bit (LSB) matching is an excellent steganographic algorithm with advantages of high payload, good visual/statistical imperceptibility and extreme ease of implementation. However, some steganalyzers take advantage of the distortion of one-dimensional histogram of LSB matching and can, to some extent, perceive the covert message. Following Cachin´s theory, this paper proposes an improvement on LSB matching by minimizing the distortion of first order statistics (i.e., one-dimensional histogram). The experiments confirm that the improvement makes the embedding procedure more secure than the conventional LSB matching.