The core workflow of FedTSA

Working on designing a federated semi-supervised and semi-asynchronous learning (FedTSA) for anomaly detection in IoT networks, which considers a more realistic semi-supervised scenario for IoT networks. A semi-asynchronous model update and staleness tolerant distribution scheme is proposed to achieve the trade-off between the round efficiency and detection performance, and the staleness of local models and the participation frequency of clients are considered to balance the contributions to the global model. In addition, a group-based aggregation function is conducted to deal with the non-IID data, and the difference transmission based on the sparse matrix is adopted to reduce the communication cost. (2022-date, submitted to journal IEEE Internet of Things Journal (IOTJ) [CCF C, SCI-Q1, IF 9.471])

Yuwei Peng (彭煜玮)
Yuwei Peng (彭煜玮)
Associate Professor

My research interests include Database Systems, Big Data of Chinese Genealogies, and Digital Watermarking.