Privacy-Preserving Misbehavior Detection in IoV using Federated Learning and Blockchain

Abstract

The fast growth of the internet of vehicles (IoV), protecting data privacy and providing credible misbehavior detection have become major issues. Classical detection methods based on centralized servers are more prone to single points of failure, scalability bottlenecks, and privacy breaches. This project presents a privacy-preserving misbehavior detection system for IoV by combining federated learning (FL) and Blockchain technologies. In our solution, cars train machine learning models on their own driving data, and only model updates, instead of raw data are exchanged between the network, with user privacy ensured. The updates are hashed and stored securely on InterPlanetary FileSystem (IPFS) and registered on the Ethereum blockchain through smart contracts, making data immutable and transparent. The blockchain element, deployed through Web3.py and Ganache, ensures trustworthiness and tamper-proofing of misbehavior reports. Decentralized architecture, which does not only improve the robustness and security of the detection process but also avoids dependency on a central authority, making the system scalable and resilient.