MTL-UE: Learning to Learn Nothing for Multi-Task Learning

2025-05-08

AI Summary

Existing unlearnable example (UE) methods target single-task learning (STL) and fail to address emerging threats in multi-task learning (MTL), where adversaries exploit multi-task data to train universal or foundation models. This paper introduces MTL-UE—the first generative UE framework explicitly designed for MTL. Its core innovations include: (i) incorporating label priors and class-level feature embeddings; (ii) designing inter-task and intra-task embedding regularization to enhance robustness against cross-task adversarial perturbations; and (iii) natively supporting dense prediction tasks with plug-and-play compatibility. Extensive experiments across four MTL benchmarks, three UE baselines, five model backbones, and five task-weighting strategies demonstrate consistent and significant improvements over prior approaches, validating MTL-UE’s effectiveness and generalizability.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Most existing unlearnable strategies focus on preventing unauthorized users from training single-task learning (STL) models with personal data. Nevertheless, the paradigm has recently shifted towards multi-task data and multi-task learning (MTL), targeting generalist and foundation models that can handle multiple tasks simultaneously. Despite their growing importance, MTL data and models have been largely neglected while pursuing unlearnable strategies. This paper presents MTL-UE, the first unified framework for generating unlearnable examples for multi-task data and MTL models. Instead of optimizing perturbations for each sample, we design a generator-based structure that introduces label priors and class-wise feature embeddings which leads to much better attacking performance. In addition, MTL-UE incorporates intra-task and inter-task embedding regularization to increase inter-class separation and suppress intra-class variance which enhances the attack robustness greatly. Furthermore, MTL-UE is versatile with good supports for dense prediction tasks in MTL. It is also plug-and-play allowing integrating existing surrogate-dependent unlearnable methods with little adaptation. Extensive experiments show that MTL-UE achieves superior attacking performance consistently across 4 MTL datasets, 3 base UE methods, 5 model backbones, and 5 MTL task-weighting strategies.

Problem

• Generating unlearnable examples for multi-task learning models
• Enhancing attack robustness via intra-task and inter-task regularization
• Supporting dense prediction tasks in multi-task learning frameworks

Innovation

• Generator-based structure with label priors
• Intra-task and inter-task embedding regularization
• Plug-and-play support for existing UE methods