DDoS Detection in VANETs Using a Multilayer Perceptron and Simple Blockchain-Based Trust Management for False Positive Minimization

Vehicular Ad-hoc Networks (VANETs) are critical for safety messaging in Intelligent Transportation Systems yet are highly vulnerable to Distributed Denial-of-Service (DDoS) attacks that disrupt safety-critical communications. This paper proposes a low-latency detection framework that combines a Multilayer Perceptron (MLP) classifier with a lightweight blockchain-backed reputation scoring mechanism. The framework augments traditional Quality of Service (QoS) features with historical trust values derived from direct and indirect evidence, which are recorded on a ledger. Experiments using simulated VANET data under three DDoS scenarios which is Beacon Flooding, High Power Jamming, and Asynchronous Beacon, were conducted to evaluate three MLP architectures (with 2, 3, and 4 hidden layers). Integrating reputation features consistently improves detection performance: macro F1-scores rose from 0.62-0.70 in the baseline (QoS-only) to 0.88-0.95 in the proposed configuration. Precision and recall gains were most pronounced for classes prone to misclassification in the baseline, significantly reducing false positives in benign high-load scenarios. While validation performance is high, accuracy drops on unseen datasets due to distribution shifts, highlighting the need for adaptation to real-world variability. Future work will address this gap by scaling the blockchain node deployment, evaluating the system's resilience to more advanced adversarial conditions such as Sybil attacks, and implementing real-time cross-domain testing to enhance generalization.