An efficient blind steganalysis using higher order statistics for the neighborhood difference matrix
Swagota Bera, Monisha Sharma, S. Subramanya Sikhar, Atul Kumar Dwivedi · 2016
This paper presents a blind JPEG steganalysis technique, which outperforms other JPEG steganographic schemes i.e. Jsteg, F5, Outguess and using discrete wavelet transform. The variations caused by JPEG steganography is enhanced by calculating the different DWT 2-D arrays along all the four directions. There exists a relationship among the four nearest neighbor pixels of JPEG image. This fact is explored for the improvement in detection purpose. The two step markov's process is implemented to model this difference DWT-2-D arrays so that higher order statistics can be utilized for steganalysis. In this paper, these higher order statistics are treated as features which are calculated from the calibrated and predicted difference DWT-2D arrays image coefficients. The Support Vector Machine is implemented for the classification purpose. The performance parameters are evaluated for the detection method for all the mentioned hiding techniques. It is the finding that the detection performance gets increased with the proposed scheme.