TrueFake: A Real World Case Dataset of Last Generation Fake Images also Shared on Social Networks

2025-04-29

AI Summary

Social media compression and re-encoding severely degrade AI-generated image detection performance, yet existing benchmarks lack realistic propagation distortions. Method: We introduce TrueFake, a large-scale benchmark of 600K images encompassing synthetics from state-of-the-art generators (e.g., SDXL, DALL·E 3, StyleGAN3) and their post-propagation variants distorted by Instagram, X, and Facebook pipelines. We systematically model forensic trace degradation along social dissemination chains and propose an “in-the-wild” evaluation paradigm. Using multi-scale feature analysis and noise-aware robust training, we quantify detector performance drops under real-world conditions. Contribution/Results: We reveal that mainstream detectors suffer 32–57% average accuracy decline after social propagation. We further demonstrate that contrastive learning and noise-aware fine-tuning significantly improve cross-platform generalization. TrueFake establishes the first reproducible, operationally realistic benchmark and optimization framework for industrial-grade AI-generated image detection.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Computer Vision: Adversarial Attacks & Robustness
• Application Domains: Misinformation & Fake News

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

AI-generated synthetic media are increasingly used in real-world scenarios, often with the purpose of spreading misinformation and propaganda through social media platforms, where compression and other processing can degrade fake detection cues. Currently, many forensic tools fail to account for these in-the-wild challenges. In this work, we introduce TrueFake, a large-scale benchmarking dataset of 600,000 images including top notch generative techniques and sharing via three different social networks. This dataset allows for rigorous evaluation of state-of-the-art fake image detectors under very realistic and challenging conditions. Through extensive experimentation, we analyze how social media sharing impacts detection performance, and identify current most effective detection and training strategies. Our findings highlight the need for evaluating forensic models in conditions that mirror real-world use.

Problem

• Addressing AI-generated fake images spread via social media
• Evaluating fake detection under real-world sharing conditions
• Identifying effective detection strategies for compressed images

Innovation

• Large-scale dataset with 600,000 images
• Includes top generative techniques and social sharing
• Evaluates fake detectors under realistic conditions