Shared updatable Deepfake video content supervision method and system

Abstract

The invention relates to a shared updatable Deepfake video content supervision method and system, and the method comprises the steps: S1, inputting a Deepfake video into a preprocessing module, extracting a video key frame, and intercepting a face image as a training sample; S2, extracting spatial domain and frequency domain features of the training sample, and inputting feature information into an SVM classification model for training to obtain an initial content supervision model; S3, establishing a sharing updatable strategy based on a block chain technology, and designing an incentive mechanism to collect new and effective Deepfake video data; and S4, after the number of the collected samples reaches a threshold value, performing update training on the initial model, and after update, sharing a Deepfake video content detection method for sample contributors, and waiting for next update. Sharing of the Deepfake video content detection method can be realized, the Deepfake video content detection method is continuously updated, the problem of sample imbalance of a Deepfake video data set is effectively eliminated, the problem of overfitting is solved, and the generalization ability of a content supervision model is improved.

Description

technical field [0001] The invention relates to the fields of machine learning and Internet content supervision, in particular to a method and system for sharing and updating Deepfake video content supervision. Background technique [0002] With the development of artificial intelligence technology, deep forgery technology based on deep learning is becoming more and more mature. Through multimedia tampering tools, people's faces in videos can be tampered at will, and it is almost impossible to detect with the naked eye. With the rise of new short video content dissemination methods, Deepfake videos are spreading faster and wider, and may be used to engage in activities prohibited by laws and regulations such as endangering national security and infringing on the legitimate rights and interests of others, causing adverse effects on social stability. In December 2019, the National Internet Information Office, the Ministry of Culture and Tourism, and the State Administration of...

AI Technical Summary

Problems solved by technology

The detection method based on the intra-frame difference focuses on the detailed feature differences of the face, while ignoring the context information of the deep fake video; the detection method based on the difference between video frames depends on the number of key frames extracted, if the video is too short The detection effect is greatly reduced

At the same time, due to the fact that deepfake generation technology is constantly being updated and upgraded, existing detection methods rely on specific data sets and generation algorithms. When the new Deepfake video content comes from new deepfake technology or does not contain specific data When the sample is set, the accuracy of some detection methods will be reduced or even the detection method will fail, that is, the model has an overfitting problem.

Images