Published Paper Details:

Video in the national defense, education, monitoring, entertainment and other fields have been widely used, so data security on the internet cannot be ignored. Video encryption protects the original video information and improves the security of video information. Researchers have done a lot of research on video encryption and put forward a lot of video encryption methods. Video encryption methods are mainly divided into complete encryption and partial encryption algorithm. In order to improve the generalization performance of video encryption and reduce the amount of data in video encryption, this paper proposes a video encryption on regions of interest (ROI) method based on Faster R-CNN by combining machine learning with information security. The method trains a Faster R-CNN model using the ROI dataset firstly, and then uses the model to extract ROI in the video. Different encryption algorithms are used to encrypt ROI and non-ROI in the video respectively. To overcome the shortcomings of encryption algorithms that can only be used for a specific coded video, a special video encryption method is proposed to encrypt the video with different video coding structure and has better generalization performance. Compared with the encryption method in the video coding process, this method considers the content information of the video fully and has better performance. It can be concluded through experiments that the encryption method in this paper has the characteristics of higher security and less calculation.