Pay Less Attention to Deceptive Artifacts: Robust Detection of Compressed Deepfakes on Online Social Networks

View on arXiv ← Back to list

Authors: Manyi Li, Renshuai Tao, Yufan Liu, Chuangchuang Tan, Haotong Qin, Bing Li, Yunchao Wei, Yao Zhao

Published: 2025-06-25 15:46:41+00:00

Comment: 20 pages, 10 figures

AI Summary

This paper introduces PLADA (Pay Less Attention to Deceptive Artifacts), a novel framework for robust deepfake detection on Online Social Networks, specifically addressing challenges posed by compression-induced "block effects" and limited paired data. PLADA integrates a Block Effect Eraser (B2E) with a dual-stage attention mechanism to mitigate block effects and an Open Data Aggregation (ODA) module to effectively utilize both paired and unpaired data. Extensive experiments across 26 datasets demonstrate PLADA's superior performance over state-of-the-art methods in detecting compressed deepfakes, even with sparse paired data.

Abstract

With the rapid advancement of deep learning, particularly through generative adversarial networks (GANs) and diffusion models (DMs), AI-generated images, or ``deepfakes, have become nearly indistinguishable from real ones. These images are widely shared across Online Social Networks (OSNs), raising concerns about their misuse. Existing deepfake detection methods overlook the ``block effects introduced by compression in OSNs, which obscure deepfake artifacts, and primarily focus on raw images, rarely encountered in real-world scenarios. To address these challenges, we propose PLADA (Pay Less Attention to Deceptive Artifacts), a novel framework designed to tackle the lack of paired data and the ineffective use of compressed images. PLADA consists of two core modules: Block Effect Eraser (B2E), which uses a dual-stage attention mechanism to handle block effects, and Open Data Aggregation (ODA), which processes both paired and unpaired data to improve detection. Extensive experiments across 26 datasets demonstrate that PLADA achieves a remarkable balance in deepfake detection, outperforming SoTA methods in detecting deepfakes on OSNs, even with limited paired data and compression. More importantly, this work introduces the ``block effect as a critical factor in deepfake detection, providing a robust solution for open-world scenarios. Our code is available at https://github.com/ManyiLee/PLADA.

Key findings
PLADA achieves superior deepfake detection performance on OSNs, outperforming state-of-the-art methods across 26 datasets, particularly excelling with compressed images and limited paired data. It demonstrates robust resilience to varying compression types (quality-aware and quality-agnostic scenarios) while maintaining strong performance on raw images. The work importantly identifies and addresses the 'block effect' as a critical, deceptive artifact that misleads existing deepfake detectors.
Approach
The PLADA framework addresses deepfake detection in compressed images through two core modules: a Block Effect Eraser (B2E) and an Open Data Aggregation (ODA) module. B2E employs a dual-stage attention mechanism (Residual Guidance and Coordination Guidance) to redirect the model's attention from deceptive compression-induced block effects to genuine deepfake artifacts. ODA processes both paired and unpaired data by computing aggregation centers for real/fake and compressed/raw images, maximizing the distinction between real and fake images and enhancing robustness against varying compression types.
Datasets
ForenSynths, GANGen-Detection, Ojha-test, DiffusionForensics (total of 26 datasets for evaluation)
Model(s)
UNKNOWN
Author countries
China
← Previous
Next →