New ABC-DFL Framework Enhances Battery Intelligence in EVs

Mouhamed Amine Bouchiha and his team have unveiled ABC-DFL, an Automated Byzantine-resilient Clustered Decentralized Federated Learning framework specifically designed for electric vehicle (EV) battery management. The framework utilizes an open-permissioned blockchain to eliminate reliance on centralized aggregation, which has been a critical vulnerability in existing systems. ABC-DFL incorporates a new dynamic Quorum Byzantine Fault Tolerance (QBFT) protocol and an oracle-based aggregation layer, enhancing both security and automation in the process of managing battery data in intelligent transportation systems (ITS).

A key component of ABC-DFL is the FLECA (Filtered Layered Enhanced Clustering Aggregation) protocol, which filters out malicious updates based on an adaptive threshold from each EV's reference model. This hierarchical approach allows for robust clustering and isolation of trustworthy EV groups, significantly reducing the impact of Byzantine attacks. Experimental evaluations reveal that FLECA not only matches FedProx convergence under benign conditions but also excels in adversarial scenarios, maintaining attack impact scores below 0.10. The framework's practical viability is further validated through on-chain and off-chain benchmarks, promising a transformative impact on EV battery intelligence and management.