Limits of Deepfake Detection: A Robust Estimation Viewpoint

Abstract

Deepfake detection is formulated as a hypothesis testing problem to classify an image as genuine or GAN-generated. A robust statistics view of GANs is considered to bound the error probability for various GAN implementations in terms of their performance. The bounds are further simplified using a Euclidean approximation for the low error regime. Lastly, relationships between error probability and epidemic thresholds for spreading processes in networks are established.

References in corpus (7)

• Conditional Generative Adversarial Nets
• Exposing DeepFake Videos By Detecting Face Warping Artifacts
• DeepFakes: a New Threat to Face Recognition? Assessment and Detection
• $f$-divergence Inequalities
• In Ictu Oculi: Exposing AI Generated Fake Face Videos by Detecting Eye Blinking
• Bayesian Inference of Spreading Processes on Networks
• Robust Estimation and Generative Adversarial Nets

Cited by in corpus (2)

• Deep Learning for Deepfakes Creation and Detection: A Survey
• Business (mis)Use Cases of Generative AI