Limits of Deepfake Detection: A Robust Estimation Viewpoint

By Sakshi Agarwal, L. Varshney. 2019

51 citations recorded by the source corpus.

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.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

The Creation and Detection of Deepfakes (2020, background): “For example, by nding a bound on a model’s delay can help detect real-time aacks such as [75], and determining the limits of GANs like [7] can help us devise the appropriate strategies.”
Deep Learning for Deepfakes Creation and Detection (2019, methodology): “On the other hand, Agarwal and Varshney [54] cast the GAN-based deepfake detection as a hypothesis testing problem where a statistical framework was introduced using the information-theoretic study of authentication [55].”
Limits of Detecting Text Generated by Large-Scale Language Models (2020, background): “As we had considered previously in the context of deepfake images [18], it is also of interest to understand how error probability in detection parameterizes the dynamics of information spreading processes in social networks, e.”