ML-ALERTCHAIN: AN ENSEMBLE MACHINE LEARNING AND BLOCKCHAIN-ENABLED FRAMEWORK FOR PREDICTIVE ACCIDENT ALERTING AND TRUSTWORTHY DISSEMINATION IN IOV

Abstract [English]

Internet of Vehicles (IoV) is an important part of smart transportation, but it relies on data supplied by vehicles which may be maliciously manipulated and falsely alarmed for safety. Existing reputation based frameworks like BRAVE-IOV are mainly reactive and concentrate on the past behavior of nodes rather than proactively predicting hazards. In this research, we offer ML-AlertChain, a new system that combines Ensemble Machine Learning with Blockchain Smart Contracts to provide both prediction accuracy and trustworthy broadcast of alerts. The methodology employs a soft-voting ensemble model that combines Random Forest and XGBoost at the Roadside Unit (RSU) level to assess multi-source data, such as velocity, acceleration, and inter-vehicle distance, to predict accidents with high precision. Upon detection, an Alert Validation Smart Contract (AVSC) initiates a decentralized multi-signature verification procedure across surrounding trusted nodes to store validated alerts onto an immutable permissioned ledger. Experimental assessments with SUMO and Python TraCI show that the proposed system obtains a prediction accuracy of 94.2% which is much better than single model baselines. Moreover, the framework achieves a low dissemination latency of less than 45ms and filters out 98% of false alarm injections even when 30% of the network is hostile. These findings show that ML-AlertChain provides a strong, proactive solution for safety-critical IoV environments, with high resilience to misinformation and meets the strict timeliness requirements of real-time emergency services.