Toward Robust Non-Transferable Learning: A Survey and Benchmark

2025-02-19

AI Summary

This work addresses ethical and security risks arising from unintended generalization of deep models on unauthorized or harmful data in non-transferable learning (NTL). We systematically expose the robustness deficiencies of existing NTL methods under adversarial attacks. First, we propose the first unified formalization of the NTL task. Second, we introduce NTLBench—the first open-source evaluation benchmark for NTL—covering mainstream techniques including data isolation, gradient masking, and feature disentanglement. Third, we design a multi-dimensional robustness evaluation protocol, establishing standardized assessment criteria and a vulnerability analysis framework. Experimental results demonstrate widespread failure in generalization control across current NTL methods. This work delivers the first comprehensive survey of NTL, a reproducible evaluation infrastructure, and theoretical tools to support trustworthy AI deployment.

Technology Category

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

Application Category

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

Abstract

Over the past decades, researchers have primarily focused on improving the generalization abilities of models, with limited attention given to regulating such generalization. However, the ability of models to generalize to unintended data (e.g., harmful or unauthorized data) can be exploited by malicious adversaries in unforeseen ways, potentially resulting in violations of model ethics. Non-transferable learning (NTL), a task aimed at reshaping the generalization abilities of deep learning models, was proposed to address these challenges. While numerous methods have been proposed in this field, a comprehensive review of existing progress and a thorough analysis of current limitations remain lacking. In this paper, we bridge this gap by presenting the first comprehensive survey on NTL and introducing NTLBench, the first benchmark to evaluate NTL performance and robustness within a unified framework. Specifically, we first introduce the task settings, general framework, and criteria of NTL, followed by a summary of NTL approaches. Furthermore, we emphasize the often-overlooked issue of robustness against various attacks that can destroy the non-transferable mechanism established by NTL. Experiments conducted via NTLBench verify the limitations of existing NTL methods in robustness. Finally, we discuss the practical applications of NTL, along with its future directions and associated challenges.

Problem

• Regulate deep learning model generalization abilities
• Address robustness against various attacks in NTL
• Provide comprehensive survey and benchmark for NTL

Innovation

• Non-transferable learning framework
• Robustness evaluation benchmark
• Deep learning generalization regulation