Watch out! This app can easily create fake videos
What's the story
Creating fake videos is no longer as difficult as it used to be and doesn't require access to special hardware anymore.
A mobile application called FakeApp, launched earlier this year, allows anyone with a smartphone to create AI-powered synthetic videos popularly known as "deepfakes."
It uses AI engine to swap faces in a video with those of other people, leading to potentially dangerous use cases.
Information
How do deepfakes work?
By providing a neural network multiple samples of a specific type of data, say pictures of a person, FakeApp trains it to perform functions like detecting that face in photos and replacing it someone else's face.
Negative Impact
The future of AI-based forgery
Primarily, FakeApp led to the publishing of a lot of fake porn, where users created AI-powered videos by imposing celebrity faces onto porn stars' bodies.
Celebrities became the prime target because creating deepfakes requires a lot of face pictures of an individual from various angles.
Combine deepfakes with other AI-powered tools that can synthesize voice and handwriting, and their negative impact can increase multifold.
Information
The worst is yet to come
Fake videos are synonymous with fraud, forgery, fake news, and social engineering. Potential misuses include creating a fake video of a world leader announcing a nuclear strike and hence, triggering war. Someone might also claim a real video of their wrongdoings to be fake.
Solutions
Humans countering the effects of AI forgery
First of all, it is important for people to be aware of the damage that AI algorithms can do so that they can prevent the negative implications of tools like FakeApp.
Second, there need to be legal measures that will deter people from promoting deepfakes.
These can include heavy penalties for creating, publishing, and hosting fake videos.
Solutions
Technology countering the effects of AI forgery
A possible fix can be to use blockchain technology. These distributed ledgers are decentralized, resilient against several security threats, and tied to public and private encryption keys.
Blockchains could be used to digitally sign and confirm the authenticity of a video file. And the more people add their digital signature to that video, the more weightage it will gain as a real document.
Solutions
Solving an AI-based problem with an AI-based counter-measure
If AI can learn to swap faces in a video, it can also learn to detect the telltale signs of a video being doctored.
Google engineer Supasorn Suwajanakorn is developing an AI-powered app that can spot fake videos. It automatically scans web browsers for manipulated media content.
Notably, videos are tough to perfectly fake as they involve fabricating streams of thousands of images.