JPEG steganalysis with high-dimensional features and accuracy
K. Rajasri, T. Indhumathi · 2014
The robust technique to steganalyse the jpeg images is projected. The planned steganalytic idea is consisted of three sections: Feature extraction, Bayesian ensemble classifier and analysis. In the initial section, high-dimensional feature vector is computed for every JPEG picture within a training set. Training set contains the samples of original plus stego images. In the second fraction, family unit of sub-classifiers which is trained on the feature vectors will be incorporated to formulate optimized decisions used for doubtful images. This is done via Bayesian system. Bayesian classifier is an easy probabilistic classifier. Bayesian classifier is built based on applying Baye's theorem with well-built independence assumptions. After classification, analysis is made about the classification. This is to confirm that there is no misclassification in the result. Finally accuracy of the classifier can be calculated using Markov Random Field cliques.