Detecting Malicious Artificial Congestion in Connected Cars Environment

By 2030, the global connected car market size is projected to be USD 361 billion. Automakers such as Mercedes-Benz, BMW, Peugeot, Toyota, and others are investing billions of dollars in software development, autonomous driving, and connected vehicle technology. Many companies are developing strong products to secure connected cars and their environment. However, when a vehicle becomes more connected, it also becomes much more vulnerable to intrusion and cyber-attacks. We investigate the use of classification-based algorithms called supervised machine learning techniques such as Decision Tree, Random Forest, XGBoost, AdaBoost, Support Vector Machines (SVM), Naive Bayes, and K-Nearest Neighbors (KNN) to detect attacks on vehicles’ identities. We demonstrate, through implementation, the high precision (99%) results obtained with Random Forest, XGBoost, and Support Vector Machines (SVM). However, the execution time for XGboost seems to be faster than for the other methods. Furthermore, for generalization purposes, we evaluated these techniques on synthetic data generated using Conditional Tabular GANs. Our findings indicate that most techniques exhibit low accuracy in detecting vehicle identity attacks. To address this issue, we propose employing ensemble learning to enhance detection accuracy.