"To Catch a Deepfake, You Need to Be One": Studying and Testing the Invisible Threat
- SQADays / 38
-
40 min
The perfect candidate who doesn’t exist, or how deepfakes are used in interviews.
Imagine you're conducting an interview. The candidate is perfect, confidently answering every question, but later you find out it was an entirely different person. This is the new reality, where deepfakes allow someone to impersonate another person.
In this talk, we'll step into the scammer's shoes: we'll create real-time deepfakes and try to cheat the system. Then, we'll switch to the defensive side: looking for vulnerabilities and learning how to distinguish a fake from a real person.
I'll show how modern deepfake model architectures work and where they fail. I'll also cover how neural networks work for deepfake detection. Finally, we will go through a practical checklist of deepfake indicators that you can apply in your very next interview.
This talk is for those who don't want to end up signing an offer letter to a neural network, and for those who are used to trust but verify.