Adaptive Gated Deepfake Detection for Low-Resolution and Resource-Constrained Environments

Authors: Vaishnavi Sen, Cody Laurie, Rashida Hasan

Published: 2026-09-04 16:14:57+00:00

AI Summary

This paper introduces AdaGate-DF, an adaptive gated deepfake detection framework designed for low-resolution and resource-constrained environments. It uses image-quality cues to dynamically route samples through a dual multi-exit system, allowing high-quality images to exit earlier for computational efficiency. AdaGate-DF balances detection performance, uncertainty-aware prediction, and computational efficiency across varying image qualities.

Abstract

Deepfake detection models often rely on high-quality inputs, fixed inference paths, and computationally expensive architectures, limiting their use in low-resolution and resource-constrained settings. This paper proposes AdaGate-DF, an adaptive gated deepfake detection framework that uses image-quality cues to route samples through a dual multi-exit system so high-quality images can exit earlier and save compute. We evaluated AdaGate-DF against MaD-CoRN, DefakeHop++, and ShuffleNetV2 on two benchmark datasets (Celeb-DF and FaceForensics++) under multiple configurations to test image resolution dependence and training and inference efficiency. On Celeb-DF, AdaGate-DF achieves an AUC of 0.9370, outperforming MaD-CoRN and DefakeHop++ while maintaining a low inference latency. Resolution-based testing shows consistent improvement as input resolution increases, reaching an AUC of 0.9708 at 384 by 384. The FaceForensics++ results highlight that AdaGate-DF remains effective under class imbalance, following competitive results with evaluated models. Overall, AdaGate-DF demonstrated a practical balance between detection performance, uncertainty-aware prediction, and computational efficiency for variable-quality deepfake detection.

Key findings

AdaGate-DF achieved an AUC of 0.9370 on Celeb-DF, outperforming MaD-CoRN and DefakeHop++ with low inference latency. Performance consistently improved with increased input resolution, reaching an AUC of 0.9708 at 384x384. On the imbalanced FaceForensics++ dataset, AdaGate-DF remained competitive and effective, demonstrating a practical balance between detection performance, uncertainty-aware prediction, and computational efficiency.

Approach

AdaGate-DF employs a dual-branch architecture combining RGB spatial features and DCT-based frequency features. It incorporates a multi-exit gated inference mechanism that adaptively routes samples through fast, medium, or full inference paths based on image-quality cues like resolution, blur, and sharpness. This adaptive routing optimizes computational resources by processing simpler samples more quickly while reserving deeper analysis for more challenging inputs.

Datasets

Celeb-DF, FaceForensics++ (FF++)

Model(s)

AdaGate-DF (proposed), MaD-CoRN, DefakeHop++, ShuffleNetV2

Author countries

USA

← Previous