DF-Net: The Digital Forensics Network for Image Forgery Detection

2025-03-28

AI Summary

To address the risk of public opinion manipulation via tampered images on social media, this paper proposes DF-Net, an end-to-end deep network for robust, pixel-level forgery localization. Methodologically, DF-Net introduces two key innovations: (1) it achieves unprecedented robustness against common social-media distortions—particularly JPEG compression and resizing—without requiring pre- or post-processing; and (2) it employs a multi-scale feature-fused encoder-decoder architecture enhanced with a frequency-domain-aware module and adversarial training to improve fine-grained discriminative capability. Extensive experiments demonstrate state-of-the-art performance across four mainstream benchmarks, with average detection accuracy improvements of 3.2%–7.8%. Notably, DF-Net maintains a high mIoU of 91.4% even under severe JPEG compression (quality factor QF = 50), significantly enhancing practical deployability in real-world social media environments.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Responsible Web: Social and technical mechanisms of refusal for web technologies and applications
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

The orchestrated manipulation of public opinion, particularly through manipulated images, often spread via online social networks (OSN), has become a serious threat to society. In this paper we introduce the Digital Forensics Net (DF-Net), a deep neural network for pixel-wise image forgery detection. The released model outperforms several state-of-the-art methods on four established benchmark datasets. Most notably, DF-Net's detection is robust against lossy image operations (e.g resizing, compression) as they are automatically performed by social networks.

Problem

• Detects pixel-wise image forgery in manipulated images
• Addresses threat of opinion manipulation via social networks
• Robust against lossy operations like resizing and compression

Innovation

• Deep neural network for pixel-wise detection
• Robust against lossy image operations
• Outperforms state-of-the-art on benchmarks