Effective image stego intrusion detection system using statistical footprints of the steganogram and fusion of classifiers
J. Hemalatha, M. K. Kavitha Devi, S. Geetha · International Journal of Computer Aided Engineering and Technology · 2020
Enlightening the processing record of a digital image is a significant problem for steganalysers and the forensic analysers. In the present day, the most precise steganalysis techniques are built as supervised classifiers by extracting the feature vectors from the digital media. This paper presents an ensemble classification method for effective image stego intrusion detection system on JPEG images consists of two step process. In the first step the features are engineered as higher-order statistics for blind steganalysis. In the second step ensemble classifier is used by fusing the classifiers such as support vector machine, neural networks, k-nearest neighbours. By applying the mentioned classifiers to these features, the steganogram and the clear (unadultered) carrier signals are effectively discriminated. For generating the image dataset, images are undergone with six embedding schemes with different payload. Experimental results show that the proposed approach remarkably improve the metrics such as specificity, sensitivity and accuracy of the system.