Title :
Steganalysis of Analysis-by-Synthesis Compressed Speech
Author :
Ding, Qi ; Ping, Xijian
Author_Institution :
Zhengzhou Inf. Sci. & Technol. Inst., Zhengzhou, China
Abstract :
In this paper a steganalysis method for analysis-by-synthesis (AbS) compressed speech is proposed. In the steganalysis method, the probability difference of 0 and 1 and five statistics, including sum of histogram local flatness, sum of histogram local extremum difference, sum of histogram local variance, sum of histogram characteristic function, are used as distinguishing features. SVM classifiers are employed to discriminate stego speech from cover speech. For the example of steganalysis to speech compressed by G.729 algorithm, experiment results show the proposed algorithm has good performance.
Keywords :
pattern classification; speech coding; steganography; support vector machines; G.729 algorithm; SVM classifiers; analysis-by-synthesis compressed speech; compressed speech; histogram characteristic function sum; histogram local extremum difference sum; histogram local flatness sum; histogram local variance sum; steganalysis; Classification algorithms; Codecs; Feature extraction; Histograms; Speech; Speech coding; Support vector machines; Analysis-by-synthesis; Compressed speech; Pulse position parameters; Statistics;
Conference_Titel :
Multimedia Information Networking and Security (MINES), 2010 International Conference on
Conference_Location :
Nanjing, Jiangsu
Print_ISBN :
978-1-4244-8626-7
Electronic_ISBN :
978-0-7695-4258-4
DOI :
10.1109/MINES.2010.148