Blockchain based Federated Learning for Object Detection

Citations Over Time

Abstract

Object detection based on deep learning needs to collect and centralize a large amount of training data from multiple parties, which leads to data privacy problems. Federated learning has become an effective way to conduct deep learning when training data are distributed. However, traditional federated learning still faces the challenge of security and data heterogeneity. Specifically, the central server will introduce a single point of failure. Malicious clients and statistical heterogeneity of data will affect the performance of the model. To solve the above problems, this paper proposes FedDetectionBC, a federated learning framework for object detection model training, which takes advantage of blockchain to ensure the robustness and auditability of federated learning. To deal with data heterogeneity, a new algorithm (Exchange-FedAvg) is also proposed. Experimental results that FedDetectionBC is more robust than traditional federated learning when there exist malicious clients in the system. The proposed Exchanged-FedAvg can achieve higher accuracy with fewer communication rounds in non-IID settings.

Related Papers

• → Semi-Federated Learning(2020)22 cited
• → Towards Fault-Tolerant Federated and Distributed Machine Learning(2024)1 cited
• → Blockchain-based Federated Learning with Limited Resources(2022)1 cited
• → A Survey on Decentralized Federated Learning(2023)21 cited
• → Accelerating Fair Federated Learning: Adaptive Federated Adam(2023)1 cited