IEDBased Special Detection Technology for PQ Steganography

We propose an IED(imaginary eigenvalue decomposition) based special diction algorithm for steganalysis of JPEGbased PQ(perturbed quantization) embedding. Both statistical analysis and experimental results show that the PQ embedding decreases linear dependences of rows and columns in the spatial domain. Detection experimental results illustrate the validity of the proposed scheme. By using SVM(support vector machine) as a classifier, a test database is constructed. Comparative results show that the poposed scheme attains better detection performance than other steganalysis schemes. Moreover, the proposed steganalyzer reaches a detection accuracy of over 70% for detecting PQ embedding, and has a better performance of blind detecting.