Research Article
Steganalysis of Adaptive Multiple-Rate Speech Using Parity of Pitch-Delay Value
@INPROCEEDINGS{10.1007/978-3-030-21373-2_21, author={Xiaokang Liu and Hui Tian and Jie Liu and Xiang Li and Jing Lu}, title={Steganalysis of Adaptive Multiple-Rate Speech Using Parity of Pitch-Delay Value}, proceedings={Security and Privacy in New Computing Environments. Second EAI International Conference, SPNCE 2019, Tianjin, China, April 13--14, 2019, Proceedings}, proceedings_a={SPNCE}, year={2019}, month={6}, keywords={Steganalysis Adaptive multi-rate codec Pitch delay Bayes’s theorem Support vector machine (SVM)}, doi={10.1007/978-3-030-21373-2_21} }- Xiaokang Liu
Hui Tian
Jie Liu
Xiang Li
Jing Lu
Year: 2019
Steganalysis of Adaptive Multiple-Rate Speech Using Parity of Pitch-Delay Value
SPNCE
Springer
DOI: 10.1007/978-3-030-21373-2_21
Abstract
Exploiting the fact that the pitch period parameter in speech parameter encoding is difficult to predict, a large number of steganographic strategies choose to embed secret information in the pitch period. Several detection methods for these steganography strategies based on the pitch period have also been proposed so far, but it is still a challenge to detect the steganography accurately. In this work, a new steganalysis scheme is proposed to detect pitch period based steganography, which has lower complexity and higher accuracy compared with the existing steganalysis schemes. Firstly, we regard a frame as a calculation unit within which the parity of four sub-frames can be obtained. Secondly, after filtering and merging into 14-dimensional PBP (parity Bayesian probability) features, these features are classified by the support vector machine (SVM). We evaluate the performance of the proposed strategy with numerous speech samples encoded by the adaptive multi-rate audio codec (AMR) and compare it with the state-of-the-art strategies. The experimental results illustrate that proposed method can effectively detect the pitch-delay based steganography. It is not only superior to the existing steganalysis methods in detection accuracy, but also has outstanding real-time detection performance and robustness because of its lower feature dimension and complexity.
