Robust Deepfake Detection: Mitigating Spatial Attention Drift via Calibrated Complementary Ensembles

TL;DR

This paper introduces a multi-stream ensemble framework with a degradation engine to improve deepfake detection robustness against real-world degradations, achieving state-of-the-art zero-shot generalization.

Contribution

It proposes a novel foundation-driven forensic framework combining specialized streams and calibrated voting to mitigate spatial attention drift in deepfake detection.

Findings

Achieves highly stable zero-shot generalization in deepfake detection.

Suppresses background attention drift effectively.

Secures Fourth Place in the NTIRE 2026 Challenge.

Abstract

Current deepfake detection models achieve state-of-the-art performance on pristine academic datasets but suffer severe spatial attention drift under real-world compound degradations, such as blurring and severe lossy compression. To address this vulnerability, we propose a foundation-driven forensic framework that integrates an extreme compound degradation engine with a structurally constrained, multi-stream architecture. During training, our degradation pipeline systematically destroys high-frequency artifacts, optimizing the DINOv2-Giant backbone to extract invariant geometric and semantic priors. We then process images through three specialized pathways: a Global Texture stream, a Localized Facial stream, and a Hybrid Semantic Fusion stream incorporating CLIP. Through analyzing spatial attribution via Score-CAM and feature stability using Cosine Similarity, we quantitatively demonstrate that these streams extract non-redundant, complementary feature representations and stabilize attention entropy. By aggregating these predictions via a calibrated, discretized voting mechanism, our ensemble successfully suppresses background attention drift while acting as a robust geometric anchor. Our approach yields highly stable zero-shot generalization, achieving Fourth Place in the NTIRE 2026 Robust Deepfake Detection Challenge at CVPR. Code is available at https://github.com/khoalephanminh/ntire26-deepfake-challenge.