Least-significant-digit steganography in low bitrate speech

For steganography over speech frames, least-significant-bit (LSB) approach has been one of the most important alternatives. However, its embedded capacity is quite limited, due to the low redundancy characteristic of low bitrate speech. Thus, we suggest a novel least significant digit (LSD) method in this paper, which makes full use of the frame bits to hide secret messages and provides a larger embedding capacity than the LSB method. The LSD method exploits the multiple adjacent states of frame parameters, which are produced by multiple modifications (e.g. +1, −1, +2, −2) and encoded as LSDs using a multi-ary numeration system. Beyond providing a considerable embedding capacity, the LSD method further extends the key space for key-based steganography, which can enhance the security of covert communication. The LSD method is implemented and evaluated with G.723.1 as the codec for speeches. The experiment results show that, compared with the previous LSB method, the LSD method increases around 30% of embedding capacity, and induces less distortion given the same embedding capacity.