Improving Image Steganalyser Performance using Second Order SPAM Features Extracting through Contourlet Transform
C Bala Subramanian1, J Hemalatha2, S P Balakannan3, S Geetha4
1C Bala Subramanian, Department of Information Technology, Kalasalingam Academy of Research and Education, Virudhunagar (Tamil Nadu), India.
2J Hemalatha, Department of Computer Science and Engineering, Srividya College of Engineering and Technology, Virudhunagar (Tamil Nadu), India.
3S P Balakannan, Department of Information Technology, Kalasalingam Academy of Research and Education, Virudhunagar (Tamil Nadu), India.
4S Geetha, Department of Computing Science and Engineering, Vellore Institute of Technology, Chennai (Tamil Nadu), India.
Manuscript received on 23 November 2019 | Revised Manuscript received on 17 December 2019 | Manuscript Published on 30 December 2019 | PP: 137-143 | Volume-9 Issue-1S4 December 2019 | Retrieval Number: A11051291S419/19©BEIESP | DOI: 10.35940/ijeat.A1105.1291S419
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: The major challenge posed by feature based blind steganalysers is the scheming of useful image features, which offers true existence of the stego noise rather than the natural noise in the images. Despite hundreds of features being applied in the real time implementation, only low detection accuracy could be achieved. Hence, this paper proposes a new model for detecting the stego image coupled with an examination of the task by applying a two-step process. (a) Extraction of the second order SPAM (Subtractive Pixel Adjacency Matrix) as features and the second order SPAM features of coefficients and co-occurrence matrices of sub band images from the contourlet transform. (b) Implementation of the system, based on an efficient classifier, Support Vector Machine which is capable of providing the higher detection rate than the existing classifers. Full- fledged experimentation with huge database of clean and steganogram images produced from seven steganographic schemes with varying embedding rates, and using five steganalysers were carried out in this study. The study shows that the proposed paradigm enhances the detection accuracy rate substantially and validates its efficiency with its better performance even at low embedding rates.
Keywords: Contourlet Transform, Steganalyser, Steganographic, Stego Noise.
Scope of the Article: Image Security