Unmasking Privacy: A Reproduction and Evaluation Study of Perturbation Techniques for Collaborative Filtering

Explore a 14-minute conference talk from SIGIR 2024 focused on privacy and security in collaborative filtering systems, where researchers present their reproduction and evaluation study of obfuscation-based perturbation techniques. Delve into the findings presented by Alex Martinez, Mihnea Tufis, and Ludovico Boratto as they examine various privacy protection methods in federated learning environments. Learn about the effectiveness and implications of different perturbation techniques used to maintain user privacy while preserving the utility of collaborative filtering systems.