Covert communication through JPEG-compressed images

We describe a new covert communication algorithm based on the principles of trellis coding and wavelet-based pixel-wise masking. This algorithm is capable of hiding 4059 bits of information reliably and secretly in JPEG-compressed images with dimensions 1024 x 1024. These data are combined with the imagery pixels so that they are robust against JPEG-recompression and format conversion. Experiments verify the great advantage of our algorithm in terms of image quality and bit error rate of the extracted message. Our algorithm encodes the source message consisting of binary bits with a BCH code and a trellis code into a sequence of ±1 digits that are embedded into the wavelet domain of the host image as a watermark using the pixel-wise masking approach. The robustness of the hidden message against JPEG compression is guaranteed by the great error correction capabilities of the used BCH and trellis coders, and the incorporation of pixel-wise perceptual shaping improves the image fidelity substantially.