Trustworthy authentication mechanism for video coding at an elementary stream level

Abstract

With recent advances in AI technology, it is becoming increasingly hard to identify whether video content is authentic or fake. As a consequence, users are losing trust in the information that they consume, and media outlets are challenged to fight modified or completely faked versions of their content. To overcome this problem, the Joint Video Experts Team JVET has recently developed a mechanism for trustworthy signing and authentication of coded video data at an elementary stream level. This method is designed so that it is compatible with key functionalities of the underlying video codecs like random access or scalable coding. Moreover, it can be applied within real time encoding. Thus, it is applicable in many important use cases that are not covered by existing technologies like C2PA. The method also supports a joint authentication of video and other media data such as coded audio data. In this paper, the key features and main technical aspects of this trustworthy authentication mechanism are described.