Abstract: In order to re?ect more accurately the effect of steganography on the high-order statistical characteristics of cover image,a high-order Markov chain statistical model for digital image steganography is proposed.After a comparison of the results of high-order Markov chain model constructed by several common image scanning methods,an n-order Markov chain model for digital image steganography using Hilbert scanning method is adopted.Then the effect of steganography on empirical matrix of the proposed model is studied,a statistical measure for n-order Markov chain model to evaluate the statistical security of digital image steganography is proposed,and the boundedness and equivalence to ε-secure criterion in the given conditions of this measure are proved.Experiments show the changes of empirical matrix of the proposed model in cover and stego images,and comparing with ε-secure criterion and the statistical measure of Markov chain model,the changes of cover-image's high-order statistical distribution caused by steganography are re?ected more adequately in the statistical measure of the proposed model.