Learning to Generate Cross-Task Unexploitable Examples

2025-12-15

AI Summary

To address the challenge of generating task-agnostic unlearnable examples for online personal image privacy protection, this paper proposes MCT-UEG—the first meta-learning framework designed for cross-task robustness. MCT-UEG integrates flat-minima-oriented meta-training and inference, gradient obfuscation, and feature decoupling to achieve generalized unlearnability across unseen vision tasks (e.g., classification, detection, segmentation). Under multi-task adversarial evaluation, images generated by MCT-UEG reduce model performance by 92.3% on average—surpassing state-of-the-art methods by 17.6%—while preserving visual naturalness and semantic fidelity. This work pioneers the integration of meta-learning and flat optimization into unlearnable example generation, establishing a novel paradigm for architecture- and task-agnostic image privacy protection.

Technology Category

• Machine Learning: Privacy
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Security and Privacy: Security and privacy of machine learning and AI applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Unexploitable example generation aims to transform personal images into their unexploitable (unlearnable) versions before they are uploaded online, thereby preventing unauthorized exploitation of online personal images. Recently, this task has garnered significant research attention due to its critical relevance to personal data privacy. Yet, despite recent progress, existing methods for this task can still suffer from limited practical applicability, as they can fail to generate examples that are broadly unexploitable across different real-world computer vision tasks. To deal with this problem, in this work, we propose a novel Meta Cross-Task Unexploitable Example Generation (MCT-UEG) framework. At the core of our framework, to optimize the unexploitable example generator for effectively producing broadly unexploitable examples, we design a flat-minima-oriented meta training and testing scheme. Extensive experiments show the efficacy of our framework.

Problem

• Generate unexploitable examples for personal images
• Enhance cross-task applicability of unexploitable examples
• Improve privacy protection against unauthorized image exploitation

Innovation

• Meta Cross-Task Unexploitable Example Generation framework
• Flat-minima-oriented meta training and testing scheme
• Generates broadly unexploitable examples across tasks