Hiding Secret Messages in Huffman Trees

A novel scheme for hiding information in Huffman trees is presented. Huffman trees are graphical representations of Huffman codes, which are optimal prefix-free codes generated by a simple algorithm given the distribution of the symbols. Huffman codes are used by MP3 and JPEG file formats for compressing audio and image files, but are also used for text documents and other digital media. The capacity, robustness and security of the information hiding scheme are analyzed and measured empirically. The results show that the scheme achieves a low capacity relative to the size of the cover-object and that the channel is not robust against an active adversary which knows about the channel. The security results show that an informed passive adversary's detection accuracy is greater than 90% if she has access to the symbol distribution and equivalent to pure guessing otherwise.