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

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.

Channel-Resilient Deep-Learning-Driven Device Fingerprinting Through Multiple Data Streams

“…The Experimental results indicate that the proposed approach correctly identifies the transmitter with an accuracy of up to 98% and that the classification is immune to inter-users and inter-streams interference in multi-user MIMO systems. Basha et al [26], [27] propose a new framework that leverages MIMO systems hardware capabilities to mitigate the channel effect and showed that, for Rayleigh channels, MIMO enabled blind partial channel estimation increases the testing accuracy by up to 40% when the CNN 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.…”

“…This improvement increases with the number of receiving antennas. • For flat fading channels, we expand the work in [26], [27] by leveraging MIMO capabilities to mitigate the channel effect without the need for altering the transmitted signals, nor impacting the bit error rates when using MIMO-enabled full blind channel estimation. • We show that full blind channel estimation performed by leveraging the combined capability of MIMO and STBC improves the training accuracy by 50% over Rayleigh flat fading channel when compared to SISO.…”

An Analysis of Complex-Valued CNNs for RF Data-Driven Wireless Device Classification

“…Early works on RF fingerprinting focused on automatic modulation classification and on using model based approaches to hand-craft and design features [6], [9], [10]. More recent RF fingerprinting works have shifted towards using deep learning to extract features from RF signals automatically [2], [4], [11], [12], [13], [14]. For instance, Soltani et al [13] proposed a data augmentation method for raw IQ data samples that works without needing prior knowledge about the waveform and the receiver-transmitter coordination.…”