Temporal Forensics and Anti-Forensics for Motion Compensated Video

摘要: Due to the ease with which digital information can be altered, many forensic techniques have been developed authenticate multimedia content. Similarly, a number of anti-forensic operations recently designed make forgeries undetectable by techniques. However, like manipulations they are hide, leave behind their own forensically detectable traces. As result, forger must balance trade-off between completely erasing evidence forgery and introducing new manipulation. Because investigator is typically bound constraint on probability false alarm (P_fa), also accuracy detect use anti-forensics. In this paper, we analyze interaction examining problem authenticating videos. Specifically, study adding or deleting sequence frames from video. We begin developing theoretical model fingerprints that frame deletion addition leaves behind, then improve upon video detection technique proposed Wang Farid. Next, propose an fool develop method for detecting introduce set evaluating performance game theoretic framework analyzing interplay forger. these evaluate each our techniques, identify optimal actions both investigator.