Home // CYBER 2017, The Second International Conference on Cyber-Technologies and Cyber-Systems // View article

RF Fingerprinting for 802.15.4 Devices: Combining Convolutional Neural Networks and RF-DNA

Authors:
Bernard Lebel
Louis N. Bélanger
Mohammad Amin Haji Bagheri Fard
Jean-Yves Chouinard

Keywords: RF-DNA; Wireless Security; Physical Layer; Neural Networks; Machine Learning

Abstract:
Wireless communications have traditionally relied on the content of the message for authenticating the sender. In protocols relying on the IEEE 802.15.4 standard, such as Zigbee, it is possible for an attacker with the right knowledge and tools to emit crafted packets that will be interpreted by the receiver as being properly identified and thus, inject arbitrary data. One way of protecting oneself from this type of attack is the use of radio frequency fingerprinting through a technique called Radio Frequency Distinct Native Attribute (RF-DNA). This approach has been demonstrated to be efficient for wireless devices of different models but still lacks accuracy when trying to identify a rogue device of the same model as the lawful emitter. This is even more of a challenge when attempting to conduct the fingerprinting using a low-cost yet flexible software defined radio. To address this challenge, the current work-in-progress attempts to train a convolutional neural network in order to be able to discriminate a legitimate device from a rogue device. Initial results show promising performance but a larger dataset of devices is required to be conclusive, which will be the focus of future work.

Pages: 70 to 74

Copyright: Copyright (c) IARIA, 2017

Publication date: November 12, 2017

Published in: conference

ISSN: 2519-8599

ISBN: 978-1-61208-605-7

Location: Barcelona, Spain

Dates: from November 12, 2017 to November 16, 2017