AAG-Stega: Automatic Audio Generation-based Steganography (1809.03463v1)

Abstract: Steganography, as one of the three basic information security systems, has long played an important role in safeguarding the privacy and confidentiality of data in cyberspace. Audio is one of the most common means of information transmission in our daily life. Thus it's of great practical significance to using audio as a carrier of information hiding. At present, almost all audio-based information hiding methods are based on carrier modification mode. However, this mode is equivalent to adding noise to the original signal, resulting in a difference in the statistical feature distribution of the carrier before and after steganography, which impairs the concealment of the entire system. In this paper, we propose an automatic audio generation-based steganography(AAG-Stega), which can automatically generate high-quality audio covers on the basis of the secret bits stream that needs to be embedded. In the automatic audio generation process, we reasonably encode the conditional probability distribution space of each sampling point and select the corresponding signal output according to the bitstream to realize the secret information embedding. We designed several experiments to test the proposed model from the perspectives of information imperceptibility and information hidden capacity. The experimental results show that the proposed model can guarantee high hidden capacity and concealment at the same time.

Overview of AAG-Stega: Automatic Audio Generation-based Steganography

The paper presents a novel approach to audio steganography called AAG-Stega, which employs an automatic audio generation method as opposed to traditional carrier modification techniques. The focus of this research is on embedding secret information in audio without altering the statistical distribution of the original signal, thereby enhancing both concealment and hidden capacity.

Key Contributions

AAG-Stega deviates from classical methods that typically modify an existing audio carrier, a process which often adds noise and alters statistical features, making detection easier. Instead, AAG-Stega leverages automatic audio generation to craft audio signals from secret bitstreams, utilizing Recurrent Neural Networks (RNNs) to achieve this embedding.

The methodology employs a probabilistic model where each audio sample point is conditioned on previous points, creating a cover through the careful selection of notes from a probability distribution. The use of LSTM networks with Lookback and Attention mechanisms strengthens the model's capability to generate high-quality audio while embedding the data.

Experimental Evaluation

The research rigorously tests the AAG-Stega method on two fronts: imperceptibility and hidden capacity.

1. Hidden Capacity: The paper measures how many secret bits can be embedded per unit and expresses this as the Embedding Rate (ER). This is calculated over various sizes of the Candidate Pool (CPS), reflecting the trade-off between information capacity and concealment. Results show a flexible range of embedding rates, validating the method's adaptability.
2. Imperceptibility: Objective and subjective evaluations are conducted. Objective measures involve calculating the likelihood of generated audios compared to training data, with scores indicating a close match in statistical distribution. Subjective tests involve an "A/B/X" approach with participants attempting, and largely failing, to distinguish between original and stego-audio, further confirming the efficacy in maintaining concealment.

Implications and Future Directions

This work suggests a substantial shift towards cover synthesis in steganography, where carriers are generated rather than modified. The approach promises significant advancements in the concealment system paradigm by potentially eliminating the discernible modifications introduced in traditional steganography.

Potential future developments can explore extending this methodology to other domains, such as video or real-time audio communication, and further optimizing the balance between embedding rate and concealment. Additionally, integration with other neural network architectures could further improve the adaptability and efficiency in diverse contexts.

Conclusion

AAG-Stega demonstrates a sophisticated approach to steganography through automatic audio generation, achieving promising results in both imperceptibility and capacity. This advancement opens up new pathways for research, potentially reshaping the strategies employed for secure information embedding across various media.