Research Article
Towards Efficient and Privacy-Preserving Service QoS Prediction with Federated Learning
@INPROCEEDINGS{10.1007/978-3-030-67540-0_3, author={Yilei Zhang and Xiao Zhang and Xinyuan Li}, title={Towards Efficient and Privacy-Preserving Service QoS Prediction with Federated Learning}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 16th EAI International Conference, CollaborateCom 2020, Shanghai, China, October 16--18, 2020, Proceedings, Part II}, proceedings_a={COLLABORATECOM PART 2}, year={2021}, month={1}, keywords={QoS prediction Cloud services Privacy preservation Federated learning Communication efficiency}, doi={10.1007/978-3-030-67540-0_3} }- Yilei Zhang
Xiao Zhang
Xinyuan Li
Year: 2021
Towards Efficient and Privacy-Preserving Service QoS Prediction with Federated Learning
COLLABORATECOM PART 2
Springer
DOI: 10.1007/978-3-030-67540-0_3
Abstract
Cloud computing provides many service resources that enable large-scale cloud applications composed of services to be widely adopted in many crucial domains. Quality of Service (QoS) is often used as an indicator in service selection and composition to guarantee the quality of cloud applications. To facilitate QoS-based selection and composition, previous studies have employed collaborative filtering techniques to predict unknown QoS values as a supplement to limited user-perceived QoS data. However, Collaborative modeling approaches encounter privacy issues in the practice of QoS prediction. Users may be reluctant to collaborate through sharing data. As a result, addressing privacy threats has become a key effort towards making QoS prediction methods practical. In this paper, we leverage federated learning techniques and propose a privacy-preserving QoS prediction approach to address this challenge. We further propose several efficiency improvement techniques to significantly reduce system overhead so that the prediction model can provide results quickly and timely. We conduct experiments on a large-scale real-world QoS dataset to evaluate our approach, and the experimental results show that it can make fast and accurate predictions.
