SEPARATE FEATURE BASED STEGANALYSIS FOR CALIBRATED JPEG IMAGES

The objective of the science of steganalysis is to detect the message hidden in an image. The steganalyst needs to keep his ultimate goal to detect the image which has data hidden in it and retrieve the message. This paper makes use of statistical data analysis using different feature sets with different percentage embedding, and an analysis is done with each of them. Since blind steganalysis techniques are used, the technique of calibration to retrieve an estimate of the cover image is used. The steganalytic technique is feature based as well. This is due to the fact that the features that is sensitive to the embedding changes that are employed for steganalysis. The domain used in this paper will be Discrete Cosine Transform. The feature set will be a combination of first order, second order and Markovian set of features. The performance rate is calculated by the error detection percentage of the combination of the feature sets. The extracted features are fed into a classifier which helps to distinguish between a stego and cover image. Support Vector Machine (SVM) is used as a classifier here. Principal Component Analysis (PCA) is used for feature reduction.