Video codec identification

Video content is routinely captured and distributed in digital world. It is then common to encode video content multiple times for several purposes. Indeed, typically, a sequence is firstly encoded directly on the capturing device. Then it is re-encoded after any editing step. Moreover, an additional coding step may be applied when a sequence is uploaded on a sharing website. Identifying the used codec is a topic of interest for a forensic analyst. Indeed, it can be a useful hint to detect the device used to generate the content. In this thesis, we focus on the situation of video encoded twice, e.g., where the first coding step is applied by the generating device, and the second one by a sharing website. In this scenario we propose a codec identification method that aims to identify the codec used in the first coding step, by analyzing footprints left by this codec. The analysis is performed re-encoding the sequence with an additional coding step controlled by the analyst. By studying the correlation obtained between the input and the output of this additional coding step, we can infer with high accuracy which was the first codec used.