T2UE: Generating Unlearnable Examples from Text Descriptions

2025-08-05

AI Summary

Existing unlearnable examples (UEs) methods require access to original images, thereby exposing sensitive data and creating a privacy paradox. To address this, we propose T2UE—a novel framework that generates effective UEs solely from textual descriptions, without ever accessing the original images, thereby establishing a “zero-contact data protection” paradigm. Methodologically, T2UE synergistically integrates text-to-image diffusion models with error-minimization optimization to project textual semantics into the image noise space, synthesizing highly transferable and robust adversarial perturbations. Experiments demonstrate that the generated UEs significantly degrade performance of mainstream models across cross-modal retrieval and classification tasks. Crucially, the defensive efficacy generalizes across diverse network architectures and supervised learning settings. T2UE thus provides a secure, practical, and privacy-preserving solution for data publishing in privacy-sensitive applications.

Technology Category

• Machine Learning: Privacy
• Computer Vision: Adversarial Attacks & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Security and Privacy: Data transparency and provenance
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

Large-scale pre-training frameworks like CLIP have revolutionized multimodal learning, but their reliance on web-scraped datasets, frequently containing private user data, raises serious concerns about misuse. Unlearnable Examples (UEs) have emerged as a promising countermeasure against unauthorized model training, employing carefully crafted unlearnable noise to disrupt the learning of meaningful representations from protected data. Current approaches typically generate UEs by jointly optimizing unlearnable noise for both images and their associated text descriptions (or labels). However, this optimization process is often computationally prohibitive for on-device execution, forcing reliance on external third-party services. This creates a fundamental privacy paradox: users must initially expose their data to these very services to achieve protection, thereby compromising privacy in the process. Such a contradiction has severely hindered the development of practical, scalable data protection solutions. To resolve this paradox, we introduce extbf{Text-to-Unlearnable Example (T2UE)}, a novel framework that enables users to generate UEs using only text descriptions. T2UE circumvents the need for original image data by employing a text-to-image (T2I) model to map text descriptions into the image (noise) space, combined with an error-minimization framework to produce effective unlearnable noise. Extensive experiments show that T2UE-protected data substantially degrades performance in downstream tasks (e.g., cross-modal retrieval) for state-of-the-art models. Notably, the protective effect generalizes across diverse architectures and even to supervised learning settings. Our work demonstrates the feasibility of "zero-contact data protection", where personal data can be safeguarded based solely on their textual descriptions, eliminating the need for direct data exposure.

Problem

• Preventing misuse of private data in web-scraped datasets for AI training
• Reducing computational cost of generating unlearnable examples for on-device use
• Enabling data protection without exposing original images to third parties

Innovation

• Generates unlearnable examples from text descriptions
• Uses text-to-image model for noise mapping
• Employs error-minimization for effective noise