Creating, Weaponizing, and Detecting Deep Fakes

• In Event: Deepfakes in the 2020 Presidential Election: Detecting New Attempts at Deception

Abstract

The past few years have seen a startling and troubling rise in the fake-news phenomena in which everyone from individuals to nation-sponsored entities can produce and distribute mis-information. The implications of fake news range from a mis-informed public to an existential threat to democracy, and horrific violence. At the same time, recent and rapid advances in machine learning are making it easier than ever to create sophisticated and compelling fake images, videos, and audio recordings, making the fake-news phenomena even more powerful and dangerous. These so called AI-synthesized media (popularly referred to as deep fakes) can make people say or do things they never said or did. I will describe how we can exploit biometric features of a person to detect deep fakes. The term biometric refers to: (1) the visual features distinct to a person, and (2) the mannerism of the person, i.e., habitual gestures/poses and idiosyncrasies of speaking/behaving. These biometric features are often violated by the nature of how deep fake videos are created. Additionally, I’ll provide an overview of the creation of deep fakes, vulnerability of the detection techniques towards attacks, and the challenges associated with detection at a large scale.

Author

• Shruti Agarwal