Beijing-based CTTL Terminal labs has released a platform that can identify which faces are faked by AI.
Based on the single-frame and multi-frame methods, the researchers used the characteristics of face biometrics, jitter accuracy and other characteristics, combined with multiple rounds of training models and test set technologies, to identify anomalies in images and video files, and then distinguish authenticity.
With the development of artificial intelligence and deep learning technology, some lawbreakers have also found new loopholes and used the new technology to achieve a very realistic "face change".
For example, several popular mobile apps that appeared last year, using Deepfake technology, only need to upload a photo to fake a face.
This has even forged video during elections in certain countries. From personal reputation and life to national security, all face great threats.
CTTL Terminal labs is a subsidiary of China Institute of Information and Communication and have long been committed to the exploration and research in the field of AI technology.
After multiple rounds of testing and verification by technical staff, the accuracy rate of the authentication of the pictures and videos reached 98.9%, which can provide authentication services for Internet content operators and consumers.
It can also assist governments and industry regulators to carry out AI face-changing technology forensics and identification.
CTTL Terminal labs said that it will actively cooperate with the country and governments at all levels to provide a strong technical guarantee for the prevention of malicious abuse of deep forgery technology.
