DeepRadioID: Real-Time Channel-Resilient Optimization of Deep Learning-based Radio Fingerprinting Algorithms

Abstract

Radio fingerprinting provides a reliable and energy-efficient IoT authentication strategy by leveraging the unique hardware-level imperfections imposed on the received wireless signal by the transmitter's radio circuitry. Most of existing approaches utilize handtailored protocol-specific feature extraction techniques, which can identify devices operating under a pre-defined wireless protocol only. Conversely, by mapping inputs onto a very large feature space, deep learning algorithms can be trained to fingerprint large populations of devices operating under any wireless standard.One of the most crucial challenges in radio fingerprinting is to counteract the action of the wireless channel, which decreases fingerprinting accuracy significantly by disrupting hardware impairments. On the other hand, due to their sheer size, deep learning algorithms are hardly re-trainable in real-time. Another aspect that is yet to be investigated is whether an adversary can successfully impersonate another device's fingerprint. To address these key issues, this paper proposes DeepRadioID, a system to optimize the accuracy of deep-learning-based radio fingerprinting algorithms without retraining the underlying deep learning model. The key intuition is that through the application of a carefully-optimized digital finite input response filter (FIR) at the transmitter's side, we can apply tiny modifications to the waveform to strengthen its fingerprint according to the current channel conditions. We mathematically formulate the Waveform Optimization Problem (WOP) as the problem of finding, for a given trained neural network, the optimum FIR to be used by the transmitter to improve its fingerprinting accuracy.

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

“…Sankhe et al [12] benefit from the adaptivity feature of software-defined radios and modify the transmitter chain of these radios such that their respective demodulated symbols acquire unique characteristics that make the CNN robust to channel changes (the signal unique characteristics dominate the channel changes). Restuccia et al [20], on the other hand, show that a carefullyoptimized digital finite impulse response filter (FIR) at the transmitter's side, applying tiny modifications to the waveform to strengthen its fingerprint based on current channel conditions, can improve the accuracy from about 40% to about 60% in case of training on 5 devices. However, these research attempts depend on modifying the transmitted signals by either adding artificial impairments that are immune to the channel variations, or by filtering to alter the transmitted signals to maximize the model accuracy, thereby resulting in a potential impact on the BER.…”

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

“…Sankhe et al [9] benefit from the adaptivity feature of software defined radios and modify the transmitter chain of these radios such that their respective demodulated symbols acquire unique characteristics that make the CNN robust to channel changes (the signal unique characteristics dominate the channel changes). Restuccia et al [14], on the other hand, show that a carefully-optimized digital finite input response filter (FIR) at the transmitter's side, applying tiny modifications to the waveform to strengthen its fingerprint based on current channel conditions, can improve the accuracy from about 40% to about 60% in case of training on 5 devices. However, these works depend on modifying the transmitted signals by either adding artificial impairments that are immune to the channel variations, or by filtering to alter the transmitted signals to maximize the model accuracy, thereby resulting in a potential impact on the BER.…”

“…• Communication system-oriented approaches where signal processing and communication approaches are applied at the transmitter or the receiver to remove the channel effect and enhance the deep learning model accuracy [14], [9]. • Deep learning/data-driven approaches where data augmentation [15] or new neural network architecture designs [16] are introduced to mitigate the channel effect on RF fingerprinting.…”

Zero-Bias Deep Learning for Accurate Identification of Internet of Things (IoT) Devices

“…Specific featurebased approaches focus on deriving distinctive features (a.k.a, transmitter fingerprints) from received signals [11], [12] to recognize known devices. Deep learning based approaches do not require knowing devices' radiometric characteristics and shows even higher accuracy [13], [14]. However, the challenge of applying deep learning approaches for IoT device identification lies in two aspects: unseen device recognition, and model interpretability.…”