ML-enabled Resource-constraint Intrusion Detection Framework for Vehicular Ad-Hoc Networks

Lead Author Affiliation

Computer Engineering

Lead Author Status

Undergraduate - Sophomore

Second Author Affiliation

Computer Science

Second Author Status

Masters Student

Third Author Affiliation

Computer Science

Third Author Status

Faculty Mentor

Research or Creativity Area

Engineering & Computer Science

Abstract

Vehicular Ad-Hoc Networks (VANETs) have become an integral component of contemporary vehicular technology. This technology provides advanced features like traffic and weather reports and collision prevention to create a sophisticated driving experience. Unfortunately, this infrastructure can become victim to cyber-attacks like intrusions that aim to compromise VANET operations. In this paper, we propose a novel machine learning (ML)-based resource-constraint intrusion detection framework for VANETs. The developed algorithm utilizes dimensionality reduction to reduce the operational footprint of our training data so that it is easier to train and deploy ML-enabled intrusion detectors for protecting resource-constraint VANET infrastructures. Results showcase consistent performance for protecting VANETs against intrusions while decreasing the overall operational footprint of the framework.

Purpose

Today’s vehicles come equipped with internet connectivity which allows us to access features like hands-free calling, emergency service access, and weather reports. These vehicles can also be interconnected on the road, so that they can communicate with each other, thereby forming a Vehicular Ad-Hoc Network (VANET). VANETs include inter-vehicular and inter-infrastructure communication technologies which are forecasted to be integral components in future technologies like smart cities and autonomous vehicles. VANETs provide essential awareness like forecasting traffic congestion and weather advisories. However, the progressive adoption of VANETs in today’s automotive technology makes them a prime target for cyber attacks. Network-level attacks like intrusions aim to disrupt functionality and can be an incredible risk for users, pedestrians, and other drivers. In this research, we have created a machine learning-enabled intrusion detection framework that can detect cyber attacks in VANETs with high performance. The main purpose of this research is to develop an automated detection strategy by leveraging the usage of machine learning and adapting the solution to have a small operational footprint. This way, it can be placed in resource-constraint environments like in VANETs.

Results

The main results of this work indicate the performance of the proposed solution for detecting intrusions in the context of VANETs. We can see from the results that incorporating dimensionality reduction in our framework through using Principal Component Analysis (PCA) allows us to greatly reduce the training time and the storage memory for the result training data. Despite the smaller size of the dataset, the performance of the machine learning framework is still very comparable to that of the original whole dataset. This indicates that by using PCA and machine learning, we can develop an effective intrusion detection framework for VANETs; one that has a smaller operational footprint and can be placed in resource-constraint environments.

Significance

The protection of VANETs from cyber attacks like intrusions is an open research area. Two of the main drawbacks of using machine learning for protection solutions are the increased reliance on having large datasets and requiring substantial processing power. The proposed work showcases that by using dimensionality reduction in conjunction with machine learning, we are able to develop automated intrusion detection solutions that can be placed in resource-constraint environments like VANETs. These solutions provide less operational footprints than the traditional solution, with comparable performance at detecting intrusions.

This document is currently not available here.

Share

COinS
 

ML-enabled Resource-constraint Intrusion Detection Framework for Vehicular Ad-Hoc Networks

Vehicular Ad-Hoc Networks (VANETs) have become an integral component of contemporary vehicular technology. This technology provides advanced features like traffic and weather reports and collision prevention to create a sophisticated driving experience. Unfortunately, this infrastructure can become victim to cyber-attacks like intrusions that aim to compromise VANET operations. In this paper, we propose a novel machine learning (ML)-based resource-constraint intrusion detection framework for VANETs. The developed algorithm utilizes dimensionality reduction to reduce the operational footprint of our training data so that it is easier to train and deploy ML-enabled intrusion detectors for protecting resource-constraint VANET infrastructures. Results showcase consistent performance for protecting VANETs against intrusions while decreasing the overall operational footprint of the framework.