Towards Efficient Communication and Secure Federated Recommendation System via Low-rank Training

Ngoc-Hieu Nguyen, Tuan-Anh Nguyen, Tuan Minh Nguyen, Vu Tien Hoang, Dung D. Le, Kok-Seng Wong

Keywords: recommendation system, federated learning, communication efficiency, low-rank training

TL;DR: We propose a communication efficient framework for Federated Recommendation System.

Abstract: With increasing regulatory constraints on centralized data gathering, federated recommendation systems (FedRec) have emerged as a promising solution for safeguarding user privacy. However, the deployment of FedRec introduces challenges in communication efficiency, stemming from the need to transmit neural network models between individual user devices and a central server. This study addresses prior shortcomings in efforts to enhance communication efficiency in FedRec. Common approaches have often led to issues such as computational overheads, inadvertent disclosure of sensitive user information, model specificity constraints, and compatibility issues with secure aggregation protocols. In response, we propose a novel framework inspired by Parameter-Efficient Fine-tuning. Our framework leverages the concept of adjusting lightweight trainable parameters while maintaining most parameters in a frozen state. This innovative approach yields a substantial reduction in both uplink and downlink communication overhead, all while avoiding the introduction of additional computational burdens. Critically, our framework remains fully compatible with secure aggregation protocols, including the robust use of Homomorphic Encryption. This research offers a promising avenue to address the pressing need for efficient and secure federated recommendation systems, ensuring user privacy protection in an era of stringent data regulations. Extensive experimental results and analysis on multiple datasets across various model architectures and security mechanisms validate the effectiveness of the proposed method.

Track: User Modeling and Recommendation

Submission Guidelines Scope: Yes

Submission Guidelines Blind: Yes

Submission Guidelines Format: Yes

Submission Guidelines Limit: Yes

Submission Guidelines Authorship: Yes

Student Author: No

Submission Number: 2375