WILD: a new in-the-Wild Image Linkage Dataset for synthetic image attribution

2025-04-28

AI Summary

Synthetic image provenance attribution faces significant challenges due to the rapid proliferation of generative models, heterogeneous generation techniques, and the scarcity of high-quality, open-source benchmark datasets. To address this, we introduce WILD—the first real-world-oriented synthetic image溯源 benchmark—comprising 20 mainstream generators (10 closed-set, 10 open-set), each with 1,000 images, 50% of which undergo diverse post-processing (e.g., compression, filtering, geometric transformations). We propose a novel “closed-set training + open-set in-the-wild” dual-mode paradigm to jointly model generator evolution and post-processing perturbations. Additionally, we design a multi-granularity augmentation strategy and a cross-generator distribution alignment evaluation protocol. Within a unified evaluation framework, we systematically assess seven baseline methods across closed-set/open-set identification, verification, and robustness tasks. Our benchmark significantly enhances model generalizability and deployability under complex, realistic conditions.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Synthetic image source attribution is an open challenge, with an increasing number of image generators being released yearly. The complexity and the sheer number of available generative techniques, as well as the scarcity of high-quality open source datasets of diverse nature for this task, make training and benchmarking synthetic image source attribution models very challenging. WILD is a new in-the-Wild Image Linkage Dataset designed to provide a powerful training and benchmarking tool for synthetic image attribution models. The dataset is built out of a closed set of 10 popular commercial generators, which constitutes the training base of attribution models, and an open set of 10 additional generators, simulating a real-world in-the-wild scenario. Each generator is represented by 1,000 images, for a total of 10,000 images in the closed set and 10,000 images in the open set. Half of the images are post-processed with a wide range of operators. WILD allows benchmarking attribution models in a wide range of tasks, including closed and open set identification and verification, and robust attribution with respect to post-processing and adversarial attacks. Models trained on WILD are expected to benefit from the challenging scenario represented by the dataset itself. Moreover, an assessment of seven baseline methodologies on closed and open set attribution is presented, including robustness tests with respect to post-processing.

Problem

• Addressing synthetic image source attribution challenge
• Providing diverse dataset for model training and benchmarking
• Evaluating robustness against post-processing and adversarial attacks

Innovation

• WILD dataset for synthetic image attribution
• Closed and open set generators included
• Robust against post-processing and attacks