Feature engineering impact on position falsification attacks detection in vehicular ad-hoc network

Abdelkreem, Eslam; Hussein, Sherif; Tammam, Ashraf · 2024 · Crossref

DOI: 10.1007/s10207-024-00830-2

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical security challenge of position falsification attacks in Vehicular Ad-hoc Networks (VANETs), where malicious nodes manipulate transmitted location data to compromise road safety. While machine learning (ML)-based misbehavior detection schemes (MDS) are widely used to identify such attacks, existing literature often lacks clarity regarding the specific impact of feature engineering on model performance. The authors aim to bridge this gap by identifying the most efficient features and algorithms and evaluating how feature selection and derivation influence detection accuracy. The study employs a two-phase methodology. First, a comprehensive literature survey identifies key features and ML algorithms used in high-performing MDS models. Second, a comparative experimental study is conducted using the publicly available Vehicular Reference Misbehavior Dataset (VeReMi), which simulates five types of position falsification attacks across varying traffic densities. The authors implemented six ML models using three different algorithms. These models were trained on two distinct feature sets: one comprising all raw message features from Basic Safety Messages (BSMs) without modification, and another incorporating feature engineering techniques, including feature selection and the derivation of new informative features. The performance was evaluated using standard metrics such as accuracy, precision, recall, and F1-score. The results demonstrate that feature engineering significantly enhances the performance of MDS models. Two models utilizing engineered features performed comparably to state-of-the-art existing studies in detecting Type-1 (constant position) and Type-2 (constant offset) attacks, while surpassing them in detecting other attack types. Furthermore, when the proposed models were evaluated using a separate simulation rather than a subset of the training dataset, feature engineering proved crucial in mitigating overfitting. The study found that employing feature engineering techniques increased the average accuracy of the models by 6.31% to 47%, depending on the specific algorithm used. This substantial improvement highlights that raw data alone is insufficient for optimal detection and that transforming raw data into informative, non-redundant features is essential. The significance of this work lies in its provision of a clearer understanding of how feature engineering impacts VANET security solutions. By exposing flaws in recent evaluation processes—specifically the tendency to evaluate models on subsets of the same dataset rather than independent simulations—the paper advocates for more realistic evaluation methods. The findings suggest that researchers should prioritize feature engineering techniques to develop robust, efficient, and accurate detection schemes capable of handling the dynamic and high-volume nature of VANET communications. This contributes to the broader field by offering a validated framework for improving ML-based intrusion detection systems in intelligent transportation networks.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-25; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• v2x connected vehicle