Deep fake detectors are often used for differentiating between fake and real frames. Deep fakes are famous for manipulating political information & exploiting women. With advanced technologies, one can easily convince the detector that AI-generated fake is real.
How Deep Fake Detectors Work?
Early versions of deep fakes were relatively easy to spot. AI has made deep fakes very hard to detect. Scientists have developed AI-powered software for detecting deep fakes. Unfortunately, they are yet not able to detect deep fakes accurately.
In the image above, it is seen how to fool an AI-based deep fake detector. The first image above shows that the detector works fine with ordinary news. When researchers infuse adversarial information, it becomes difficult for the detector to judge.
Video detectors carefully examine each frame of the video. It’s impossible to make a manipulated frame appear natural. Adition of adversarial news to a frame can influence the detector’s judgment. This tricks the detector into believing that the frame is normal.
The attack works on White & Black box AI. According to researchers fakes do not get distorted even when the frame is compressed.
How Dangerous is Deep Fake?
Conclusion
The team has concluded that a bad actor with little knowledge of Deep fakes can fool the detectors. They can very easily make frames and convince the viewers into what they want them to believe.
Deep fakes have limited good applications but unlimited bad applications. Exploiting people would be very easy. Unless govt. regulates the deep fake industry, trusting the news sources would be difficult for a common man.
With such high tech software, controlling false information would be very difficult. Social media sites now are taking action to stop fake news. They have introduced various checks for broadcasting news. This does not eliminate all the fake news but does some good.
