Title

Using messages precedence similarity to detect message injection in in-vehicle network

Author

Mubark JedhFollow

Degree Type

Creative Component

Semester of Graduation

Fall 2020

Department

Electrical and Computer Engineering

First Major Professor

Doug Jacobson

Second Major Professor

Lotfi Ben Othmane

Degree(s)

Master of Science (MS)

Major(s)

Information Assurance

Abstract

We developed an intrusion detection technique that detects an attack on CAN Bus network using Long Short-Term Memory Recurrent Neural Network (LSTM RNN). Modern cars contain a number of Electronic Control Units (ECUs) that are responsible for normal car functionality and safety. These ECUs communicate with each other by an in-vehicle bus communication network such as CAN Bus. The CAN Bus lacks encryption and authentication mechanism to ensure secure communication. The safety of the driver relies on the ECUs exchanging messages on the CAN Bus securely. We proposed an intrusion detection system (IDS) that secure in-vehicle communication network. We develop a technique that counts the number of precedence messages, then we use a similarity matrices to detect injection of messages into the CAN Bus. First, We develop message precedences similarity graphs that distinguish between an attack state and no attack state. Then, we constructed an LSTM RNN for anomalous detection. Then, we use similarity matrices graph to training the LSTM RNN at different size and time window. Also, we develop a method to detect changes in PIds centrality due to the injection of messages on CAN Bus. The method was not successful in detecting attacks in-vehicle communication network.

Copyright Owner

Mubark B Jedh

Copyright Year

2020

File Format

PDF

Embargo Period (admin only)

9-21-2020

Recommended Citation

Jedh, Mubark, "Using messages precedence similarity to detect message injection in in-vehicle network" (2020). Creative Components. 651.
https://lib.dr.iastate.edu/creativecomponents/651

1