With the rapid development of digital technology, the problem of media tampering has become increasingly serious, and higher education teaching resources are facing a crisis from image tampering and sound tampering to video tampering and even deep forgery, which brings great challenges to education work. Through deep learning, tampering can be automatically detected and self-learning and optimized in the process of detection, so that it can remain efficient and accurate in the face of new tampering techniques. The purpose of this paper is to discuss in detail the application of deep learning in media tampering detection and to explore its application in the protection of higher education teaching resources.