EEG-Features for Generalized Deepfake Detection

TL;DR

This work explores whether human EEG responses can aid Deepfake detection and generalize beyond trained manipulation methods. By presenting FaceForensics++ stimuli to a participant and decoding real versus fake faces with an SVC on EEG features, the study demonstrates not only in-domain detection but also out-of-domain generalization to unseen Deepfake types. Two processing variants (V1 using PCA and V2 using spatial–temporal chunking) show above-chance performance and statistically significant results, suggesting a generalized neural representation of artificiality in computer-generated faces. If validated with more participants and stimuli, this human-in-the-loop approach could complement CNN detectors and improve robustness to novel Deepfake artifacts, potentially informing the design of more realistic digital avatars while clarifying how digital realism is processed cognitively.

Abstract

Since the advent of Deepfakes in digital media, the development of robust and reliable detection mechanism is urgently called for. In this study, we explore a novel approach to Deepfake detection by utilizing electroencephalography (EEG) measured from the neural processing of a human participant who viewed and categorized Deepfake stimuli from the FaceForensics++ datset. These measurements serve as input features to a binary support vector classifier, trained to discriminate between real and manipulated facial images. We examine whether EEG data can inform Deepfake detection and also if it can provide a generalized representation capable of identifying Deepfakes beyond the training domain. Our preliminary results indicate that human neural processing signals can be successfully integrated into Deepfake detection frameworks and hint at the potential for a generalized neural representation of artifacts in computer generated faces. Moreover, our study provides next steps towards the understanding of how digital realism is embedded in the human cognitive system, possibly enabling the development of more realistic digital avatars in the future.

Figures (1)

• Figure 1: Mean EEG responses with respect to the three stimuli classes for electrode PO8 as well as the topography (across all electrodes) of the difference between fake and real images at 385 ms after stimulus onset (green box).