Loading...

Universal steganalysis based on local prediction error in wavelet domain

Shojaei Hashemi, A ; Sharif University of Technology

749 Viewed
  1. Type of Document: Article
  2. DOI: 10.1109/IIHMSP.2011.95
  3. Abstract:
  4. A passive universal image steganalysis method is proposed that is shown to be of higher detection accuracy than existing truly blind steganalysis methods including Farid's and the WAM. This is achieved by improving some weaknesses of Farid's steganalysis scheme in feature extraction, that is, instead of deriving an over-determined equation system for each sub band of the wavelet decomposition, the sub bands are divided into overlapping blocks and an over-determined equation system is constructed for each block. To guarantee the existence of finite answers, the over-determined equation systems are solved in a way different from Farid's by using Moore-Penrose pseudo-inverse concept. Further improvement to the performance is achieved by adding diagonal directions and increasing the number of moments. The comparative evaluations confirm the superiority of the proposed method over prevalent blind steganalysis schemes
  5. Keywords:
  6. support vector machine ; Blind steganalysis ; Comparative evaluations ; Detection accuracy ; Equation systems ; Local prediction ; Moore-Penrose ; Pseudo-inverses ; Steganalysis ; Sub-bands ; Support vector ; Universal steganalysis ; Wavelet domain ; Error detection ; Feature extraction ; Multimedia signal processing ; Steganography ; Wavelet decomposition
  7. Source: Proceedings - 7th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIHMSP 2011 ; 2011 , Pages 165-168 ; 9780769545172 (ISBN)
  8. URL: http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6079560