Efficient Least-significant-bits Steganography for VoIP

For the instantaneity of real-time applications such as voice over Internet protocol (VoIP), the processing delay and efficiency need to be carefully considered. Matrix coding is a widely used coding scheme for steganography to improve the embedding efficiency which means the average hiding bits per covering bit change. In order to take full advantage of it, a classifying least-significant-bits (LSB) with frame window (CLFW) scheme is proposed for low bitrate speech covers in VoIP communication. The classifying method selects LSB with similar characteristics which bring consistent effect to speech quality for every bit modification. Meanwhile, the frame window strategy collects more speech frames in VoIP speech stream for matrix coding, which enhances the embedding efficiency. In CLFW, frame LSB are classified into three levels to construct separate sets of covering bits, which limits the range of embedding modification. As a result, the speech quality degradation caused by feature discrepancies of speech frames is minimized. The experimental results demonstrate that CLFW method largely increases the embedding efficiency compared with existing schemes and is much more imperceptible than direct matrix coding and simple LSB steganography.