Exposing GAN-generated Faces Using Inconsistent Corneal Specular Highlights

TL;DR

This paper presents a method to detect GAN-generated faces by analyzing inconsistencies in corneal specular highlights between the two eyes, exploiting a common artifact caused by the lack of physical constraints in GANs.

Contribution

The authors introduce an automatic technique to extract and compare corneal highlights, revealing a new artifact for identifying GAN-synthesized faces.

Findings

Inconsistent corneal highlights are prevalent in high-quality GAN faces.

The proposed method effectively distinguishes real from GAN-generated faces.

Corneal highlight analysis provides a simple yet powerful detection tool.

Abstract

Sophisticated generative adversary network (GAN) models are now able to synthesize highly realistic human faces that are difficult to discern from real ones visually. In this work, we show that GAN synthesized faces can be exposed with the inconsistent corneal specular highlights between two eyes. The inconsistency is caused by the lack of physical/physiological constraints in the GAN models. We show that such artifacts exist widely in high-quality GAN synthesized faces and further describe an automatic method to extract and compare corneal specular highlights from two eyes. Qualitative and quantitative evaluations of our method suggest its simplicity and effectiveness in distinguishing GAN synthesized faces.