Title :
Single Band Statistics and Steganalysis Performance
Author :
Le, Pham Hai Dang ; Franz, Matthias O.
Author_Institution :
Inst. for Opt. Syst. (IOS), HTWG Konstanz Univ. of Appl. Sci., Konstanz, Germany
Abstract :
Most universal steganalysis techniques use an image model to reconstruct an estimate of the original, unmanipulated cover from the input. Differences between reconstructed and input images are an indication of a steganographic manipulation. In this paper, we analyze the relation between the modeling error of the image model and detection performance in the wavelet domain. Based on the modeling error we define a measure of separability which is highly correlated with detection performance. We find that in uncompressed images only fine scales play a role in steganalysis, whereas separability is spread out over all scales in JPEG images.
Keywords :
image coding; image reconstruction; statistical analysis; steganography; wavelet transforms; JPEG images; detection performance; image model; image reconstruction; single band statistics; steganalysis performance; steganalysis techniques; steganographic manipulation; uncompressed images; wavelet domain; Image coding; Image color analysis; Predictive models; Transform coding; Wavelet domain; Wavelet transforms; steganalysis; steganography; universal steganalysis;
Conference_Titel :
Intelligent Information Hiding and Multimedia Signal Processing (IIH-MSP), 2010 Sixth International Conference on
Conference_Location :
Darmstadt
Print_ISBN :
978-1-4244-8378-5
Electronic_ISBN :
978-0-7695-4222-5
DOI :
10.1109/IIHMSP.2010.54
