Title (eng): Signal intrusion detection for remote keyless entry systems

Author: Hasler, S. (Simon D.)

Description (deu): Masterarbeit, Fachhochschule St. Pölten, Masterstudiengang Information Security, 2018

Description (eng): Recent years have brought to light a surprising number of hacking techniques that circumvent the security measures implemented in automobiles, allowing car thieves to remotely open vehicles without the use of the legit keyfob. White-hat hackers and security researchers have revealed how these kinds of attacks are possible and what kind of hardware and software is used. In this research work, I review the RollJam attack, which aims at replaying captured signals after preventing the car from receiving them by jamming the receiver frequency during legit transmissions. I show that this attack scheme can be reproduced to remotely unlock a 2008 model VW Group vehicle with a selection of low-cost transmitter devices and open-source software. After visualizing captured signals from different transmitters and analyzing their unique characteristics, I proceed by demonstrating that a number of features can be extracted that allow to distinguish between signals based on their origin. Based on my findings, I present a technique that applies two different machine learning algorithms for the classification of data points on a pre-built dataset, and subsequently use it to create a proof-of-concept for a Signal Intrusion Detection System capable of classifying unknown signals based on known signal data. I show how both machine learning algorithms perform in various use cases on the provided signal data in terms of resource utilization and accuracy, and reveal where their individual strength and weaknesses lie. Lastly, I introduce the nVidia Jetson TX2 module that I chose as the hardware platform for the tested proof-of-concept and explain why it is especially well-suited for AI computing tasks in embedded environments such as automobiles.

Object languages: English

Date: 2018

Rights: © All rights reserved

Classification: Kraftfahrzeugdiebstahl ; Hacker

Permanent Identifier