Multisource traffic incident reporting and evidence management in Internet of Vehicles using machine learning and blockchain

Intelligent transportation systems require efficient ways to manage traffic incidents like accidents and traffic rule violations. Secure logging of incident related data is important in case of accidents for forensic analysis, insurance, and legal settlements. Internet of vehicles can be enhanced using machine learning techniques to detect events automatically at the vehicle and road infrastructure level and blockchain can be used as a source for immutable evidence storage and management. The work proposes a framework addressing multiple challenges in traffic incident detection and evidence management like gathering of traffic event related evidence from multiple sources, addressing fault management in embedded vehicle on board units, resiliency from malicious event reporting and handling conflicting traffic incident event reports. The framework relies on a combination of cooperative event correlation and trust model to detect malicious and erroneous reporting of traffic incidents followed by Long Short term Memory (LSTM) and Bayesian model to resolve conflicting event reports. The events are correlated and verified before being transmitting onto the blockchain for evidence management and access control by various stakeholders. Analysis is presented highlighting the improvement in efficiency achieved by convergence of the multiple approaches proposed.