Unlearning Comparator: A Visual Analytics System for Comparative Evaluation of Machine Unlearning Methods

TL;DR

This paper presents Unlearning Comparator, a visual analytics system designed to systematically evaluate and compare machine unlearning methods, focusing on accuracy, efficiency, and privacy aspects through model comparison and attack simulation.

Contribution

It introduces a novel visual analytics tool that enables detailed analysis and comparison of MU methods, addressing the challenge of understanding their trade-offs.

Findings

Helps users understand model behavior changes after unlearning

Enables simulation of membership inference attacks for privacy evaluation

Facilitates systematic comparison of MU methods at multiple levels

Abstract

Machine Unlearning (MU) aims to remove target training data from a trained model so that the removed data no longer influences the model's behavior, fulfilling "right to be forgotten" obligations under data privacy laws. Yet, we observe that researchers in this rapidly emerging field face challenges in analyzing and understanding the behavior of different MU methods, especially in terms of three fundamental principles in MU: accuracy, efficiency, and privacy. Consequently, they often rely on aggregate metrics and ad-hoc evaluations, making it difficult to accurately assess the trade-offs between methods. To fill this gap, we introduce a visual analytics system, Unlearning Comparator, designed to facilitate the systematic evaluation of MU methods. Our system supports two important tasks in the evaluation process: model comparison and attack simulation. First, it allows the user to compare the behaviors of two models, such as a model generated by a certain method and a retrained baseline, at class-, instance-, and layer-levels to better understand the changes made after unlearning. Second, our system simulates membership inference attacks (MIAs) to evaluate the privacy of a method, where an attacker attempts to determine whether specific data samples were part of the original training set. We evaluate our system through a case study visually analyzing prominent MU methods and demonstrate that it helps the user not only understand model behaviors but also gain insights that can inform the improvement of MU methods. The source code is publicly available at https://github.com/gnueaj/Machine-Unlearning-Comparator.