Detecting Localized Deepfakes: How Well Do Synthetic Image Detectors Handle Inpainting?

TL;DR

This paper evaluates how well existing deepfake detectors, trained on fully synthetic images, can identify localized inpainting manipulations across various datasets and techniques, revealing partial transferability and detection strengths.

Contribution

It provides a systematic assessment of the generalization of deepfake detectors to localized inpainting, highlighting their capabilities and limitations in cybersecurity contexts.

Findings

Models trained on multiple generators partially transfer to inpainting detection.

Detectors reliably identify medium- and large-area inpainting manipulations.

Existing detectors outperform some ad hoc approaches in localized deepfake detection.

Abstract

The rapid progress of generative AI has enabled highly realistic image manipulations, including inpainting and region-level editing. These approaches preserve most of the original visual context and are increasingly exploited in cybersecurity-relevant threat scenarios. While numerous detectors have been proposed for identifying fully synthetic images, their ability to generalize to localized manipulations remains insufficiently characterized. This work presents a systematic evaluation of state-of-the-art detectors, originally trained for the deepfake detection on fully synthetic images, when applied to a distinct challenge: localized inpainting detection. The study leverages multiple datasets spanning diverse generators, mask sizes, and inpainting techniques. Our experiments show that models trained on a large set of generators exhibit partial transferability to inpainting-based edits and can reliably detect medium- and large-area manipulations or regeneration-style inpainting, outperforming many existing ad hoc detection approaches.