Deep Federated Learning Based on Knowledge Distillation and Differential Privacy
Hui Lin, Feng Yu, Xiaoding Wang · 2023
The Internet of Things (IoT) sensors have become prevalent in numerous fields, including industrial production, smart homes, environmental protection, medical diagnosis, and bioengineering. While efficient data fusion enhances the quality of intelligent services provided by IoT, the process of data fusion carries the risk of privacy leakage due to the sensitive nature of the perceived data. In this chapter, we propose a deep federated learning algorithm that utilizes knowledge distillation and differential privacy to safeguard privacy during the data fusion process. Our approach involves adding Gaussian noise at different stages of knowledge distillation-based federated learning to ensure privacy protection. Our experimental results demonstrate that this strategy provides better privacy preservation while achieving high-precision IoT data fusion.