Audio steganography by modification of cepstrum at a pair of frequencies

A method of embedding information in the cepstral domain of a cover audio signal is proposed. An earlier technique that combines the psychoacoustical masking property with the decorrelation property of speech cepstrum for audio steganography has been extended to achieve faster and stronger key-based embedding implementation without compromising imperceptibility, payload, or data retrieval. Use of a pair of arbitrary frequencies in the proposed method in place of auditorily masked frequencies, which form the key for embedding and retrieval, shows comparable results in imperceptibility of stego and payload with less computational effort and potentially more keys available. Accurate data recovery in the presence of low levels of noise or bandpass filtering over telephone bandwidth demonstrate the usefulness of the proposed technique in data hiding and covert communication applications.