Byzantine Robust Federated Learning (BRFL) was jointly proposed in October 2023 by a research team from universities and institutions such as Beijing University of Aeronautics and Astronautics and Guangxi Normal University. The relevant research results were published in the paper "...".BRFL: A Blockchain-based Byzantine-Robust Federated Learning Model".

Byzantine Robust Federated Learning (BRFL) comprises two main components: the Pearson Correlation Consensus Algorithm (PPCC) and the Precision-Based Spectrum Aggregation (PSA) algorithm. PPCC selects the aggregation node for the next round based on the Pearson correlation coefficient between the local model and the global model from previous rounds, while simultaneously validating the accuracy of the local model using the local dataset of the aggregation node, addressing the lack of test datasets in federated learning. PSA clusters highly correlated local models and verifies their accuracy by calculating the average, thereby detecting malicious models and addressing resource cost issues. Experimental results demonstrate that BRFL exhibits high robustness and effectively reduces resource consumption.