Title :
Steganalysis in Multiple Sources of Cover Images
Author_Institution :
Software Coll., Shenyang Normal Univ., Shenyang, China
Abstract :
45 dimensional feature vectors were extracted from DCT and spatial domains of images to defeat steganographic schemes in this paper. Compared with commonly used steganalytic algorithms, the presented method offers much better performances in double JPEG compressed covers and their stegos. However, when our method was applied to detect whether multiple sources of images containing hidden data or not, it showed very poor performances. An improved training technique was proposed and a robuster steganalytic algorithm was designed.
Keywords :
data compression; discrete cosine transforms; feature extraction; image classification; image coding; steganography; support vector machines; DCT; JPEG image; cover image; discrete cosine transform; feature vector extraction; hidden data; spatial domain; steganalysis; steganographic scheme; support vector machine classifier; training technique; Algorithm design and analysis; Discrete cosine transforms; Feature extraction; Histograms; Pixel; Robustness; Statistical distributions; Statistics; Steganography; Transform coding; Steganalysis; feature vector; robust;
Conference_Titel :
Hybrid Intelligent Systems, 2009. HIS '09. Ninth International Conference on
Conference_Location :
Shenyang
Print_ISBN :
978-0-7695-3745-0
DOI :
10.1109/HIS.2009.50