Putting spotlight on the plaguing phenomenon of ‘DeepFake’ in #SMLive 2019, Sagar Kaul, Founding CEO, MetaFact says we are seeing more of a negative side of it than the positive.
Kaul’s company is developing an AI-based tool to detect misinformation and disinformation in real time. He kickstarted the session by clearing the concept of Deep Fake – a technique for human image synthesis based on artificial intelligence. It is used to combine and superimpose existing images and videos onto source images or videos using a machine learning technique, known as generative adversarial network.
Application of DeepFake – India and Gobal
Kaul shared that the concept got pace when the superimposing of images of Hollywood celebrities on porn stars started flooding in the West. At the end of 2017, one of Taylor Swift’s morphed videos made headlines; that’s when people realized that manipulation of images can be done in this manner.
It created a stir when the technique was applied to morph an image of Former US President Barack Obama – of a politician of such grandeur- it was the best example to show then that how easy it is to create DeepFake video.
“The whole idea of DeepFake is to create disharmony, spread fake news and misinformation,” he added.
In India also, a female journalist’s image was morphed with a porn star;s and the video was circulated through WhatsApp.
How does ‘DeepFake work?
While explaining how the technology performs, Kaul stated that the program or application basically takes those photos, understands facial features and tries to impose those facial features onto video wherein we want the face swap to be done. Synthesize the voice-its easier to do that both. Creating a DeepFake would take you 2-5 mins.
How do we spot deepfake
“To create a technology that is able to understand and detect deep fakes in real time is a priority. But the problem is that technology at our disposal is okay but who are creating DeepFake they too are looking out the same way and can create loopholes in them”.
Other ways you can spot the difference is through your own eyes. He briefed, “Since we use images to morph a video, pulled out form internet or Facebook, in majority of the photos we always have our eyes open because nobody likes to get clicked with their eyes closed. So, in a DeepFake video, you will always see minimum eye click- The video is doctored, that is the first sign”.
“When a person speaks- the moment of the body might be natural but at the same time, there would be no blinks. Always look at the ends of the face. If an image is manipulated and put on a morphed video, pixelations are at the end of your face”.
How to fight DeepFakes
Kaul highlighted here that technology is being researched. If there is any video being morphed, respective apps will pull them down. “But you have to be updated with the latest findings. Not only DeepFake but whatever you find on any other platforms. It is also advisable to do your research online,” he asserts.
Kaul added that if you see someone posting that on any platform- just take an initiative. Send a simple message I think what you have shared is incorrect.
Until we educate ourselves and each other- the problem is not going to be solved.
It is a problem that can be of huge risk for the security of your country and yourself.