A Method for Collecting Vehicular Network Data for Enhanced Anomaly Detection

Samuel De La Motte, Jin B. Hong

Research output: Chapter in Book/Conference paperConference paperpeer-review

Abstract

Advances in anomaly and intrusion detection systems (IDS) for vehicle networks can enable attacks to be stealthier, such as mimicking the behaviours of aggressive driving to disrupt normal vehicle operations. Moreover, data collection, classification and evaluation has been challenging due to vehicle manufacturers deploying proprietary encoding schemes for sensor communications. This paper proposes a method for vehicle network data collection and processing to provide naturalistic datasets for identifying different driving behaviours. Firstly, we provide a guide for the assembly of an Arduino Uno R3 with a Sparkfun CAN bus shield for connection to the on-board diagnostics (OBD) port to collect the vehicle data. Secondly, we offer baseline metrics for labelling passive and aggressive driving behaviours through the analysis of a vehicle’s CAN bus data. Finally, a state-of-the-art deep learning model that combines the convolutional neural network (CNN) and recurrent neural network (RNN) architectures is implemented to evaluate the fitness of the data collected using the proposed method. The collected naturalistic dataset was used to train this model in-order to recognise driver behaviour patterns that are otherwise unobservable to humans and that can be used to identify passive and aggressive driver events. Our experiment shows a 0.98 F1 score, indicating our data collection method and metrics used for selecting features can provide quality datasets for intrusion detection against attacks mimicking aggressive driving.

Original languageEnglish
Title of host publicationInformation Security Applications - 22nd International Conference, WISA 2021, Revised Selected Papers
EditorsHyoungshick Kim
PublisherSpringer Science + Business Media
Pages16-27
Number of pages12
ISBN (Print)9783030894313
DOIs
Publication statusPublished - 2021
Event22nd World Conference on Information Security Application, WISA 2021 - Jeju, Korea, Republic of
Duration: 11 Aug 202113 Aug 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13009 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference22nd World Conference on Information Security Application, WISA 2021
Country/TerritoryKorea, Republic of
CityJeju
Period11/08/2113/08/21

Fingerprint

Dive into the research topics of 'A Method for Collecting Vehicular Network Data for Enhanced Anomaly Detection'. Together they form a unique fingerprint.

Cite this