Classifying WiFi "physical fingerprints" using complex deep learning

• NCJ Number: 305718
• Author(s): Logan Smith; Nicholas Smith; Joshua Hopkins; Daniel Rayborn; John E. Ball; Bo Tang; Maxwell Young
• Date Published: April 2020
• Grant Number(s): 2018-75-CX-K002
• Agencies: NIJ-Sponsored
• Publication Type: Research (Applied/Empirical)
• Award: RF Fingerprinting for Contraband Wireless Devices Identification, Detection and Tracking in Correctional Facilities.

Annotation

This conference presentation lays out four new deep neural networks for classifying WiFi physical fingerprints, in attempts to address WiFi susceptibility to security breaches by adversarial actors mimicking Media Access Controller addresses of currently connected devices.

Abstract

In this conference presentation, the researchers describe four new deep neural networks (NNs) for classifying WiFi physical fingerprints: a real-valued deep NN; a corresponding complex-valued deep NN; a real-valued deep convolutional NN (CNN); and the corresponding complex-valued deep CNN. These efforts attempt to prevent the problem of WiFi susceptibility to security breaches by adversarial actors who mimic Media Access Controller (MAC) addresses of currently connected devices by classifying devices according to their physical fingerprints due to the fact that fingerprints are unique for each device as well as independent of MAC addresses. Results show state-of-the-art performance against a dataset of nine WiFi devices.