System-Aware Unlearning Algorithms: Use Lesser, Forget Faster

2025-06-06

AI Summary

This work addresses machine unlearning under system constraints: minimizing model degradation while approximating retraining performance and defending against system-aware adversaries upon fulfilling user deletion requests. We propose *system-aware unlearning*, a novel paradigm that relaxes the traditional worst-case adversarial assumption by providing guarantees solely based on data accessible within the system. We formally define this paradigm for the first time and characterize a quantitative trade-off between storage overhead and unlearning efficiency. We design an exact linear classifier unlearning algorithm based on selective sampling and generalize it to broad function classes. Furthermore, we establish theoretical bounds linking unlearning capability to key system resources—deletion capacity, accuracy, memory, and runtime. Experiments demonstrate that our method achieves unlearning quality comparable to full retraining, while significantly reducing both memory footprint and computational cost.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Game Theory and Economic Paradigms: Adversarial Learning
• Multiagent Systems: Adversarial Agents

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Machine unlearning addresses the problem of updating a machine learning model/system trained on a dataset $S$ so that the influence of a set of deletion requests $U subseteq S$ on the unlearned model is minimized. The gold standard definition of unlearning demands that the updated model, after deletion, be nearly identical to the model obtained by retraining. This definition is designed for a worst-case attacker (one who can recover not only the unlearned model but also the remaining data samples, i.e., $S setminus U$). Such a stringent definition has made developing efficient unlearning algorithms challenging. However, such strong attackers are also unrealistic. In this work, we propose a new definition, system-aware unlearning, which aims to provide unlearning guarantees against an attacker that can at best only gain access to the data stored in the system for learning/unlearning requests and not all of $Ssetminus U$. With this new definition, we use the simple intuition that if a system can store less to make its learning/unlearning updates, it can be more secure and update more efficiently against a system-aware attacker. Towards that end, we present an exact system-aware unlearning algorithm for linear classification using a selective sampling-based approach, and we generalize the method for classification with general function classes. We theoretically analyze the tradeoffs between deletion capacity, accuracy, memory, and computation time.

Problem

• Minimize influence of deleted data on machine learning models
• Develop efficient unlearning algorithms against realistic attackers
• Balance deletion capacity, accuracy, memory, and computation time

Innovation

• System-aware unlearning for realistic attackers
• Selective sampling for efficient linear classification
• Tradeoffs analysis for deletion capacity and accuracy