64-bit quantization: taking payload capacity of speech steganography to the limits

A fundamental problem in steganography is the trade-off between the hiding capacity and the quality of the stego signal. Researchers have found in the Discrete Cosine Transform (DCT) a potential solution to this trade-off challenge due to its strong compaction property. The DCT has the ability to represent a time-domain signal within a few numbers of DCT coefficients leaving a substantial amount of insignificant DCT coefficients that can host the secret data. Moreover, it has been found that the compaction property becomes stronger with highly-correlated signals. Therefore, this paper suggests to fully exploit the strong compaction property of the DCT by segmenting the cover speech signal into correlated segments. Hiding is performed in the insignificant DCT coefficients in each segment which are determined by a new quantization method suitable for speech signals. Experimental results have proven that the proposed steganography technique has outperformed competing speech hiding schemes, in terms of the payload capacity which reached up to 14.94 bps, at similar or higher stego quality.

Automated summary

A trade-off exists between hiding capacity and stego signal quality in steganography. This paper proposes utilizing the strong compaction property of DCT by segmenting the cover speech signal and hiding the secret data in the insignificant DCT coefficients. Experimental results demonstrate that the proposed technique outperforms competing schemes in terms of payload capacity and stego quality.