Detection of Synthetic Face Images: Accuracy, Robustness, Generalization

TL;DR

This paper presents a study on detecting synthetic face images, showing high accuracy with simple models trained on specific generators, but highlighting challenges in generalization and robustness against adversarial attacks.

Contribution

The study introduces FF5, a dataset of synthetic face images from multiple generators, and demonstrates that simple models can effectively detect known generators but struggle with unseen ones.

Findings

High accuracy in detecting known synthetic images

Vulnerability to adversarial attacks

Poor generalization to unseen generators

Abstract

An experimental study on detecting synthetic face images is presented. We collected a dataset, called FF5, of five fake face image generators, including recent diffusion models. We find that a simple model trained on a specific image generator can achieve near-perfect accuracy in separating synthetic and real images. The model handles common image distortions (reduced resolution, compression) by using data augmentation. Moreover, partial manipulations, where synthetic images are blended into real ones by inpainting, are identified and the area of the manipulation is localized by a simple model of YOLO architecture. However, the model turned out to be vulnerable to adversarial attacks and does not generalize to unseen generators. Failure to generalize to detect images produced by a newer generator also occurs for recent state-of-the-art methods, which we tested on Realistic Vision, a fine-tuned version of StabilityAI's Stable Diffusion image generator.