On the Security of Image Manipulation Forensics

In this paper, we present a unified understanding on the formal performance evaluation for image manipulation forensics techniques. With hypothesis testing model, security is qualified as the difficulty for defeating an existing forensics system and making it generate two types of forensic errors, i.e., missing and false alarm detection. We point out that the security on false alarm risk, which is rarely addressed in current literatures, is equally significant for evaluating the performance of manipulation forensics techniques. With a case study on resampling-based composition forensics detector, both qualitative analyses and experimental results verify the correctness and rationality of our understanding on manipulation forensics security.

• Type: Journal
• Year: 2015
• Where: PCM
• Authors: Gang Cao, Yongbin Wang, Yao Zhao, Rongrong Ni, Chunyu Lin
• Keywords: Multimedia, PCM 2015