Digital Video Forensics Algorithm Based On Spatial And Temporal Matching

Recently with the rapid development of information technology, a variety of multimedia video editing software is coming up,such as Photoshop, Premiere Pro, Adobe After Effects,Mokey and so on. The operation of the video editing software is very simple and the effect is significant. The person without a rich experience in video codec and video editing can also do it. Because of the popularity of these video editing software, the network is filled with a lot of tampering videos. How to determine the authenticity and integrity of the video resources has become a hot topic in the field of information security.Copy-paste is a typical pattern of tampering to delete an object in a digital video, so it is widely used for the detection of tampering. In this paper, we deeply analyze BBC Motion Key plug-in of Adobe After Effects. We can move the object in video sequences by the software. Firstly it estimates the movement by use of time-domain correlation characteristics and optical flow techniques. And then it selects the appropriate area from frames before and after to fill the current patch to delete the corresponding region of the target object.We propose an algorithm based on spatial and temporal matching by analyzing the tampering mechanism of Adobe After Effects. We first find the best matching unit in the spatial domain, and then select the best matching unit in temporal domain using a different matching method. After re-coding quantization noise will be exported to the video sequence. Therefore, we use quantization noise factor to fix the time domain matching criteria which reduces the influence of lossy compression coding. The algorithm can detect the tampering area and adapt to the impact of lossy compression coding system. Finally we can accurately determine whether the test video sequence is tampered or not.