Identify the Copy & Paste Frames in Compressed Surveillance Video Based on Stream Parameters

Frame duplication is a typical method of tampering surveillance video which replacing a video clips with the repetitive copies of one single frame to pretend a static scene in order to conceal some important events. The forensics algorithms based on the image content can hardly identify the forged clip from the video sequence of the static scene because of the high similarity of the frames in a surveillance video. While the tampering trace could be found in the parameters of the compressed video stream because the codec is quite sensitive to the slight different between frames even it's invisible. By investigating the stream parameters, the code rate and the residual are revealed to be the prominent parameters related to such tampering. This paper analyzes the characteristics about the stream parameters, and discusses the behaviors of the code rate and the residual in the tampered video. Then an algorithm is designed to identify the duplication of frames based on the values of bit rate and residual. The indicators based on the bit rate and the residual is constructed and a fused criterion is proposed to identify the tampered frames. The experimental results show that the proposed method can identify effectively the duplication for the surveillance within an acceptable running time.