Leveraging Multiple Transmissions and Receptions for Channel-Agnostic Deep Learning-Based Network Device Classification

TL;DR

This paper introduces a MIMO-based approach to improve wireless device classification accuracy by mitigating channel effects, achieving significant gains in both AWGN and Rayleigh channels, especially across different channel conditions.

Contribution

It is the first to leverage MIMO capabilities for channel-agnostic device classification, significantly enhancing robustness against channel variations in RF fingerprinting.

Findings

Combining multiple received signals improves accuracy by up to 30% in AWGN channels.

Blind channel estimation with MIMO increases accuracy by up to 40% in Rayleigh channels when training and testing on the same channel.

Accuracy improves by up to 60% when testing on different channels from training.

Abstract

The accurate identification of wireless devices is critical for enabling automated network access monitoring and authenticated data communication in large-scale networks; e.g., IoT. RF fingerprinting has emerged as a solution for device identification by leveraging the transmitter unique manufacturing impairments. Although deep learning is proven efficient in classifying devices based on the hardware impairments fingerprints, DL models perform poorly due to channel variations. That is, although training and testing neural networks using data generated during the same period achieve reliable classification, testing them on data generated at different times degrades the accuracy substantially, an already well recognized problem within the community. To the best of our knowledge, we are the first to propose to leverage MIMO capabilities to mitigate the channel effect and provide a channel-resilient device classification. We show that for AWGN channels, combining multiple received signals improves the testing accuracy by up to $30\%$. We also show that for Rayleigh channels, blind channel estimation enabled by MIMO increases the testing accuracy by up to $40\%$ when the models are trained and tested over the same channel, and by up to $60\%$ when the models are tested on a channel that is different from that used for training.