Reproducibility Companion Paper: Making Users Indistinguishable: Attribute-wise Unlearning in Recommender Systems

TL;DR

The paper tackles protecting sensitive user attributes in recommender systems by evaluating Post-Training Attribute Unlearning (PoT-AU) with a two-component loss that balances recommendation performance and unlearning. It introduces two distinguishability measures, User-to-User (U2U) and Distribution-to-Distribution (D2D), and demonstrates through reproduction on multiple datasets that U2U-R and D2D-R can effectively reduce attacker success while preserving recommendation quality. The work provides comprehensive artifacts—including datasets, preprocessing scripts, configuration files, and code—to facilitate reproducibility and cross-study comparisons in privacy-preserving unlearning for recommender systems. This contributes to the practical impact of privacy-aware machine learning by enabling rigorous, repeatable evaluations and broader adoption of unlearning techniques in real-world systems.

Abstract

In this paper, we reproduce the experimental results presented in our previous work titled "Making Users Indistinguishable: Attribute-wise Unlearning in Recommender Systems," which was published in the proceedings of the 31st ACM International Conference on Multimedia. This paper aims to validate the effectiveness of our proposed method and help others reproduce our experimental results. We provide detailed descriptions of our preprocessed datasets, source code structure, configuration file settings, experimental environment, and reproduced experimental results.