Abstract | ||
---|---|---|
Steganography and Steganalysis have attracted a lot of attention in decades. Recently, voice communication has been more and more popular, which provides ways to covert communication. However, the existing audio steganalysis methods can only gain good detection accuracies when the hidden ratio is high. Besides, majority of the audio steganalysis methods can not provide a general evaluation, only provide the detection accuracies according to several high hidden ratios. In this paper, we proposed a new method for audio steganalysis by introducing linear prediction method, a technique from signal coding and speaker identification filed, into audio steganalysis, which can bring significant differences between covers and stegos. The linear prediction based features are utilized as the classification features loaded in a support vector machine for detection. In our work we used hidden message to cover ratio to replace the concept of hidden ratio, providing a uniform criterion to compare the performance among steganalysis methods. Furthermore, we exploited a general dataset, in which the hidden message size ranges from several bits to the maximum hiding capacity for a general evaluation on steganalysis methods. Experiment results show that our method delivers a better performance than previous two prestigious methods and brings above 96% accuracy. In general evaluation, our method gains a higher score than the other two methods. Steganalysis is a challenging work, this linear prediction based method maybe an approach to bring improvement to this filed and provide inspiration for other form of media steganalysis. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1007/s11042-017-5123-x | Multimedia Tools Appl. |
Keywords | Field | DocType |
Linear prediction, Audio steganalysis, Steganography, Support vector machine | Steganography,Speaker identification,Signal coding,Pattern recognition,Computer science,Voice communication,Support vector machine,Speech recognition,Linear prediction,Artificial intelligence,Steganalysis,Message size | Journal |
Volume | Issue | ISSN |
77 | 12 | 1380-7501 |
Citations | PageRank | References |
2 | 0.39 | 15 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Chunling Han | 1 | 2 | 0.39 |
Rui Xue | 2 | 119 | 18.19 |
Rui Zhang | 3 | 3 | 0.76 |
Xueqing Wang | 4 | 119 | 6.45 |