The deceptive world of deepfakes

Consider this: You receive a call from a colleague to transfer a big sum for an apparently-legitimate purpose. The voice on the other side is a person you know well. What would you do?
To transfer the amount can be a big mistake. Because, it can be an AI-assisted deep-voice attack.
 
And, it’s not all fiction. Something similar happened in 2020 when a Hong Kong-based banker received a call from the director of his company to transfer $35 million for an acquisition. It was a computer simulation of the director’s voice he was talking to. The elaborate fraud came with fake emails to verify the purchase. 
 
In a world of growing cybersecurity threats, Deepfakes are possibly the most dangerous yet fascinating among the lot. It is a portmanteau of ‘deep learning’ and ‘fake’. With the technology, one can create realistic videos of, say a celebrity or even a friend, doing and saying things they have not done or said.
 
That’s why it’s not surprising to find Vladimir Putin appearing on your screen to declare peace, and Volodymyr Zelenskyy telling his people to surrender. The deepfakes are taken down, but they reveal the role of misinformation in the ongoing Russia-Ukraine crisis.
 
Unconvincing fakes of Putin and Zelenskyy was ridiculed by many, but often the realistic nature of such content makes it hard to fact-check. This makes synthetic media a powerful tool that can be weaponized for malicious purposes.

Individuals have unique faces, voices, figures, expressions, speech patterns and movements. Taking all that into consideration, AI can analyse video footage of an individual to extrapolate how they would speak or do something.
In general, Deepfake programs use Generative Adversarial Networks that have two algorithms. While one forges Deepfakes, the other identifies flaws in the forgery, which are corrected subsequently.

The technology has disrupted politics, mocked TV shows, targeted influential people, generated blackmail materials, created internet memes and satires – thereby making way to mainstream consciousness.
And this has also inspired researchers around the world to develop technologies to detect Deepfakes.
 
For instance, Microsoft has announced technologies that can detect manipulated content and assure people that the media they’re viewing is authentic.
And it’s not limited to videos only. With sinister audio Deepfakes, you can’t believe your ears either. However, this emerging domain can have applications in computer game designs, medical applications and more.

Watch video