Meta-Learning and Targeted Differential Privacy to Improve the Accuracy-Privacy Trade-off in Recommendations
For developers of privacy-preserving recommender systems, this work offers a practical method to reduce accuracy loss from differential privacy while maintaining privacy guarantees.
The paper proposes combining targeted differential privacy (applied only to stereotypical user data) with meta-learning to improve the accuracy-privacy trade-off in recommender systems, achieving better accuracy and lower empirical privacy risk than uniform DP and full DP baselines.
Balancing differential privacy (DP) with recommendation accuracy is a key challenge in privacy-preserving recommender systems, since DP-noise degrades accuracy. We address this trade-off at both the data and model levels. At the data level, we apply DP only to the most stereotypical user data likely to reveal sensitive attributes, such as gender or age, to reduce unnecessary perturbation; we refer to this as targeted DP. At the model level, we use meta-learning to improve robustness to remaining DP-noise. This achieves a better trade-off between accuracy and privacy than standard approaches: Meta-learning improves accuracy and targeted DP leads to lower empirical privacy risk compared to uniformly applied DP and full DP baselines. Overall, our findings show that selectively applying DP at the data level together with meta-learning at the model level can effectively balance recommendation accuracy and user privacy.