Abstract
Steganalysis of adaptive multi-rate (AMR) speech is a hot topic for controlling cybercrimes grounded in steganography in related speech streams. In this paper, we first present a novel AMR steganalysis model, which utilizes extreme gradient boosting (XGBoost) as the classifier, instead of support vector machines (SVM) adopted in the previous schemes. Compared with the SVM-based model, this new model can facilitate the excavation of potential information from the high-dimensional features and can avoid overfitting. Moreover, to further strengthen the preceding features based on the statistical characteristics of pulse pairs, we present the convergence feature based on the Markov chain to reflect the global characterization of pulse pairs, which is essentially the final state of the Markov transition matrix. Combining the convergence feature with the preceding features, we propose an XGBoost-based steganalysis scheme for AMR speech streams. Finally, we conducted a series of experiments to assess our presented scheme and compared it with previous schemes. The experimental results demonstrate that the proposed scheme is feasible, and can provide better performance in terms of detecting the existing steganography methods based on AMR speech streams.
Highlights
Steganography is a security technique of embedding secret information into a certain carrier while the secret information can be extracted accurately
We mainly focus on the steganalysis on adaptive multi-rate (AMR) speech streams, which has a wide application in Voice over IP (VoIP) streams
In this work, motivated by the above analysis, we present an AMR-based speech steganalysis based on extreme gradient boosting (XGBoost) [38,39,40,41,42,43]
Summary
Steganography is a security technique of embedding secret information into a certain carrier while the secret information can be extracted accurately. Taking the AMR speech codec at 12.2 kbps mode as an example, η usually has three setting values (1, 2, 4), which correspond to different steganographic bandwidths Their experiments demonstrated that, by selecting an optimal η, this method could achieve a good balance between speech quality and embedding capacity. We propose a convergence feature for detecting the FCB-based steganography, which describes the global characteristics of AMR speech streams. Mainly reflects local characteristics of AMR speech streams, the presented convergence feature is a useful complement to SCPP. XGBoost rather than traditional SVMs employed in previous schemes as the classifier It enjoys some advantages, such as mining the potential information from the hybrid features, avoiding overfitting (making assumptions too strict to get consistent ones) and having a strong generalization ability. A conclusion about our work is presented
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have