Investigation of Feature Engineering Methods for Identifying Attacks in the VANET

This article discusses the problem of increasing the efficiency of machine learning methods in identifying attacks in the VANET network by expanding the feature space using Feature Engineering methods. The main idea of this work is to generate new features of a dataset using pre-trained models such as support vector machines for classification and Kmeans for clustering. The analysis of the efficiency of the generated features was carried out when solving the problem of identifying attacks using such machine learning methods as KNN, Random Forest, XGB, CatBoost, LGBM. Computational experiments showed that when SVM-based features were included, most ensemble machine learning methods improved accuracy by an average of 0.137% while adding a cluster number based on Kmeans resulted in an average efficiency improvement of 0.493%. At the same time, for the studied machine learning methods, the performance decreased slightly.