What is it about?

Recommender systems utilize users' data to generate recommendations even though this harms users' privacy, since the data can be exposed through the recommendations. Existing approaches add random noise to the data to ensure differential privacy, but this leads to a drop in recommendation quality. Therefore, we propose our neighborhood reuse approach that improves the trade-off between recommendation quality and privacy.

Featured Image

Why is it important?

Our approach reuses users' data for generating recommendations and in this way, only a small set of users needs to be protected with differential privacy. Most users' data is not utilized for generating recommendations and therefore, does not need to be protected. With this, our approach can reduce the amount of noise that is incorporated into the recommendation process, and overall, this leads to better user privacy and higher recommendation quality than comparable approaches.


We propose our novel neighborhood reuse principle which improves the trade-off between recommendation quality and user privacy in recommender systems. This tackles an important problem for users and recommender systems providers.

Peter Müllner
Know-Center GmbH

Read the Original

This page is a summary of: ReuseKNN: Neighborhood Reuse for Differentially-Private KNN-Based Recommendations, ACM Transactions on Intelligent Systems and Technology, July 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3608481.
You can read the full text:




The following have contributed to this page