In February, OpenAI released videos created by its generative artificial intelligence program Sora. The strikingly realistic content, produced via simple text prompts, is the latest breakthrough for companies demonstrating the capabilities of AI technology. It also raised concerns about generative AI's potential to enable the creation of misleading and deceiving content on a massive scale.
The lab's tools use a sophisticated machine learning program called a constrained neural network. This algorithm can learn, in ways similar to the human brain, what is"normal" and what is"unusual" at the sub-pixel level of images and videos, rather than searching for specific predetermined identifiers of manipulation from the outset.
"Until now, forensic detection programs have been effective against edited videos by simply treating them as a series of images and applying the same detection process," Stamm said."But with AI-generated video, there is no evidence of image manipulation frame-to-frame, so for a detection program to be effective it will need to be able to identify new traces left behind by the way generative-AI programs construct their videos.
For the test, the group trained eight CNN detectors, including MISLnet, with the same test dataset used to train the image detectors, which including real videos and AI-generated videos produced by the four publicly available programs. Then they tested the program against a set of videos that included a number created by generative AI programs that are not yet publicly available: Sora, Pika and VideoCrafter-v2.