Reverb hiding: A new framework for audio steganography

Abstract

In the last decades, a lot of new approaches have been developed for audio steganography. However, most of these studies ignore musical hearing. In this paper, we proposed a new framework for hiding data in audio using the reverb effect, which is one of the time-dependent effects in music production. Reverb is an audio effect in music production that creates an artificial atmosphere that simulates what could be found in a natural setting such as a large hall, stadium, etc. The impulse response of the related environment is convolved with an audio file in a studio environment. This study proposes embedding secret data in audio signals with the reverb effect. The secret message is embedded into the impulse response of the natural environment and then the audio signal is reverbed with this impulse response. The original and reverberated audios are again blended for parallel processing, resulting in stego-audio. In the data extraction stage, firstly the fully reverberated audio is obtained from stego-audio which was blended with the original (dry) audio. Then, the reverberated audio file is deconvolved with the dry audio again, and the stego impulse response is obtained.

Existing studies of audio steganography such as echo hiding are common techniques used in music production but have a significant impact on hearing. The proposed reverb hiding is done with much more complex parameters, so the stego-audio is more compatible with real-life applications. The reverb-hiding achieves superior PSNR results for different sizes of the secret message. The proposed method evaluated in terms of oscillograms, frequency spectrum analysis, spectrograms, cepstrogram analysis, histogram analysis, and autocorrelation analysis. The experimental results are also compared with six different existing audio-tech methods. The proposed reverb-hiding method significantly impacts the trade-off between capacity and imperceptibility, as indicated by the results.