Influence of Speech Codecs Selection on Transcoding Steganography

The typical approach to steganography is to compress the covert data in order to limit its size, which is reasonable in the context of a limited steganographic bandwidth. TranSteg (Trancoding Steganography) is a new IP telephony steganographic method that was recently proposed that offers high steganographic bandwidth while retaining good voice quality. In TranSteg, compression of the overt data is used to make space for the steganogram. In this paper we focus on analyzing the influence of the selection of speech codecs on hidden transmission performance, that is, which codecs would be the most advantageous ones for TranSteg. Therefore, by considering the codecs which are currently most popular for IP telephony we aim to find out which codecs should be chosen for transcoding to minimize the negative influence on voice quality while maximizing the obtained steganographic bandwidth.

💡 Research Summary

The paper investigates how the choice of speech codecs influences the performance of TranSteg, a transcoding‑based steganographic method designed for IP telephony. Unlike traditional steganography, which merely compresses covert data, TranSteg deliberately compresses the overt voice payload by transcoding it from a higher‑bit‑rate codec to a lower‑bit‑rate one, thereby creating free bits that can carry hidden information. This approach promises a much larger steganographic bandwidth, but it also risks degrading voice quality because the transcoding process can introduce distortion, especially in the high‑frequency range. The authors therefore set out to identify which codec pairs provide the best trade‑off between hidden‑channel capacity and perceived speech quality.

To that end, six codecs that are widely deployed in modern VoIP systems—G.711, G.722, G.726, G.729, iLBC, and Speex—are selected. Each codec is examined both as an overt (original) codec and as a covert (target) codec, yielding a matrix of possible transcoding combinations. The study proceeds in two phases. First, the theoretical steganographic capacity of each pair is estimated by calculating the bit‑rate difference before and after transcoding; this difference directly translates into the amount of payload that can be embedded per second. Second, the authors conduct extensive subjective and objective quality assessments. They use a diverse set of speech samples, apply the transcoding, embed hidden data, and then evaluate the resulting speech with Mean Opinion Score (MOS) and PESQ metrics. Additional measurements include packet loss, end‑to‑end latency, and CPU utilization to gauge the feasibility of real‑time deployment.

Results reveal a clear pattern: pairs with a large compression‑ratio gap yield the highest hidden bandwidth, while the impact on quality depends heavily on the specific codecs involved. The G.711 (64 kbps) → G.729 (8 kbps) combination provides roughly 48 kbps of steganographic capacity and maintains an average MOS of 3.6, which is considered “good” in telephony. The G.722 (64 kbps) → iLBC (13.33 kbps) pair offers slightly lower capacity but achieves MOS scores near 4.0, indicating almost transparent quality. Conversely, combinations such as G.726 → Speex, where the compression gain is modest and the target codec introduces significant spectral distortion, result in MOS values below 2.8 and are deemed impractical.

Latency analysis shows that moving from a high‑bit‑rate to a low‑bit‑rate codec generally reduces decoding delay, benefiting real‑time communication, whereas computationally intensive codecs like Speex can increase processing time and cause noticeable jitter. Consequently, the optimal codec selection must consider network bandwidth, acceptable latency, and the desired steganographic throughput.

The authors conclude that TranSteg can achieve substantially higher covert bandwidth than conventional methods without sacrificing acceptable voice quality, provided that the codec pair is chosen wisely. The most promising configurations identified are G.711 → G.729, G.722 → iLBC, and G.711 → iLBC. They suggest future work on adaptive transcoding algorithms that react to network conditions, multi‑stream simultaneous embedding, and robust error‑correction schemes to mitigate packet loss.