The School of Computer Science is pleased to present…
Date: Tuesday, August 13th, 2024
Time: 10:00 AM- 11:30 AM
Location: Memorial Hall Room 109
Vertical federated learning enables parties, owning different features of the same sample space to train a machine learning model collaboratively while retaining their data locally. Each party keeps both its data and model local but exchanges intermediate computed results. However, the revealed intermediate results are vulnerable to various inference attacks. Recently, many solutions targeting vertical federated learning take advantage of cryptography, trusted execution environment, or differential privacy to provide privacy to the parties. However, these solutions incur computation and communication overhead limiting the participation of low-resourced devices in the training process or affect the model’s accuracy. We present FedMod, an innovative privacy-preserving vertical federated learning approach that utilizes multi-server secret sharing. Experimental results demonstrate that FedMod can achieve comparable performance to traditional centralized machine learning while providing the added benefits of federated learning, such as data privacy and security.
Internal Reader: Dr. Alioune Ngom
External Reader: Dr. Mohammad Hassanzadeh
Advisor: Dr. Dima Alhadidi
