NoT: Federated Unlearning via Weight Negation

2025-03-07

AI Summary

In federated learning, efficiently enabling participant removal without accessing raw data or auxiliary storage poses a critical privacy-compliance challenge. This paper proposes the first weight-inversion (×−1)-based federated unlearning mechanism, theoretically proven to disrupt inter-layer co-adaptation while ensuring both thorough unlearning and efficient model recovery. Our method requires neither historical gradients, local models, nor the to-be-removed data; it achieves unlearning via a single round of lightweight perturbation and trajectory analysis, making it natively compatible with mainstream federated frameworks. Experiments across three benchmark datasets and three model architectures demonstrate that our approach improves unlearning success rate by 12.6%–28.4%, reduces communication overhead by 57%–73%, and cuts computational cost by 41%–69% over state-of-the-art baselines—substantially outperforming existing solutions.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Search and Optimization: Learning to Search
• Natural Language Processing: Safety and Robustness

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Federated unlearning (FU) aims to remove a participant's data contributions from a trained federated learning (FL) model, ensuring privacy and regulatory compliance. Traditional FU methods often depend on auxiliary storage on either the client or server side or require direct access to the data targeted for removal-a dependency that may not be feasible if the data is no longer available. To overcome these limitations, we propose NoT, a novel and efficient FU algorithm based on weight negation (multiplying by -1), which circumvents the need for additional storage and access to the target data. We argue that effective and efficient unlearning can be achieved by perturbing model parameters away from the set of optimal parameters, yet being well-positioned for quick re-optimization. This technique, though seemingly contradictory, is theoretically grounded: we prove that the weight negation perturbation effectively disrupts inter-layer co-adaptation, inducing unlearning while preserving an approximate optimality property, thereby enabling rapid recovery. Experimental results across three datasets and three model architectures demonstrate that NoT significantly outperforms existing baselines in unlearning efficacy as well as in communication and computational efficiency.

Problem

• Federated unlearning removes participant data from FL models.
• NoT algorithm uses weight negation for efficient unlearning.
• NoT outperforms baselines in efficacy and efficiency.

Innovation

• Weight negation for federated unlearning
• No additional storage or data access needed
• Perturbs model parameters for quick re-optimization