Synthetic Attack Dataset Generation With ID2T for AI-Based Intrusion Detection in Industrial V2I Network

Industrial Vehicle-to-Infrastructure (iV2I) networks are increasingly adopted in settings such as warehouses, construction sites, and smart factories to enhance automation and operational efficiency. However, these systems face growing cybersecurity risks that threaten safety-critical operations. This paper introduces a realistic synthetic dataset created using the ID2T framework, which injects malicious traffic, such as DDoS, PortScan, and memory corruption exploits, into benign communication traces collected from actual iV2I environments. The resulting hybrid dataset, combining synthetic and real-world traffic, enables the supervised training of a Multi-Layer Perceptron (MLP) neural network using 16 meticulously crafted flow-based features. Experimental results demonstrate high detection accuracy under both balanced and threat-specific conditions, validating the effectiveness of ID2T in modeling domain-relevant cyberattack behaviors. In addition to strong classification performance, this work demonstrates how synthetic malicious traffic generation reduces the cost and complexity of cyberattack emulation. The proposed method offers a scalable and reproducible framework for training intrusion detection systems (IDS), highlighting the critical role of Artificial Intelligence (AI) in securing next-generation industrial vehicular networks.