Selecting Cover for Image Steganography by Correlation Coefficient

The security of steganographic system is improved by selecting cover. The cover data are modeled as Gauss-Markov process, where the correlation coefficient of two arbitrary data elements is the exponent of correlation parameter. The KL divergence and Bhattacharyya distance of Spread Spectrum steganographic system increase with the correlation parameter. Thus the cover with smaller correlation parameter is selected to improve security. For image spatial domain steganography, the correlation parameter of image data is calculated by a special exponential model of correlation coefficients and the least squares estimator. Experiments show that the cover selection method is efficient on improving the security of image steganoagraphy.