Abstract: Based on the principle of blind detection techniques, a data hiding scheme that modifies high frequency integer-wavelet coefficients is proposed. Coefficients of a subband above a wavelet denoising threshold are chosen for data embedding. To preserve the histogram of wavelet coefficients, an adjustment strategy is introduced around the threshold. Since most blind detection methods select classification features according to
differences of statistical distributions between the cover and stego images, the proposed method can resist blind detection. Experimental results show that the method outperforms previous steganographic methods such as LSB matching and pixel-value differencing in terms of the capability of resisting current typical universal blind detecting methods.