Real-time detecting one specific tampering operation in multiple operator chains

摘要

Currently, many forensic techniques have been developed to determine the processing history of given multimedia contents. However, because of the interaction among tampering operations, there are still fundamental limits on the determination of tampering order and type. Up to now, a few works consider the cases where multiple operation types are involved in. In these cases, we not only need to consider the interplay of operation order, but also should quantify the detectability of one specific operation. In this paper, we propose an efficient information theoretical framework to solve this problem. Specially, we analyze the operation detection problem from the perspective of set partitioning and detection theory. Then, under certain detectors, we present the information framework to contrast the detected hypotheses and true hypotheses. Some constraint criterions are designed to improve the detection performance of an operation. In addition, Maximum-Likelihood Estimation (MLE) is used to obtain the best detector. Finally, a multiple chain set is examined in this paper, where three efficient detection methods have been proposed and the effectiveness of our framework has been demonstrated by simulations.