Considerations for Radio Frequency Fingerprinting across Multiple Frequency Channels

Abstract

Radio Frequency Fingerprinting (RFF) is often proposed as an authentication mechanism for wireless device security, but application of existing techniques in multi-channel scenarios is limited because prior models were created and evaluated using bursts from a single frequency channel without considering the effects of multi-channel operation. Our research evaluated the multi-channel performance of four single-channel models with increasing complexity, to include a simple discriminant analysis model and three neural networks. Performance characterization using the multi-class Matthews Correlation Coefficient (MCC) revealed that using frequency channels other than those used to train the models can lead to a deterioration in performance from MCC > 0.9 (excellent) down to MCC < 0.05 (random guess), indicating that single-channel models may not maintain performance across all channels used by the transmitter in realistic operation. We proposed a training data selection technique to create multi-channel models which outperform single-channel models, improving the cross-channel average MCC from 0.657 to 0.957 and achieving frequency channel-agnostic performance. When evaluated in the presence of noise, multi-channel discriminant analysis models showed reduced performance, but multi-channel neural networks maintained or surpassed single-channel neural network model performance, indicating additional robustness of multi-channel neural networks in the presence of noise.

RF eigenfingerprints, an Efficient RF Fingerprinting Method in IoT Context

“…Since then, several authors have explored the RF fingerprinting method for IoT using feature learning approaches [6,30,31]. Particularly, many works address the complexity problem in IoT context using a lightweight method [32][33][34]. For example, in [32], the authors present an lightweight procedure based on mobile edge computing (MEC) in IoT context.…”

“…For example, in [32], the authors present an lightweight procedure based on mobile edge computing (MEC) in IoT context. Furthermore, some authors mentioned that another important aspect of RF fingerprinting for IoT is scalability [5,33], i.e., the capacity of the algorithm to be retrained easily. Indeed, the authors mentioned that Siamese network architectures can solve this problematic, especially using one-shot learning.…”

“…In [21], the authors used 720,000 examples to train their network, i.e., 180,000 examples per class. In [5,33], the authors suggested that Siamese networks could be used to solve this problem of scalability. Indeed, Siamese networks are able to perform one-shot learning, i.e., learn a class representation using a single example.…”

“…The final noise n (t) is an additive white Gaussian complex noise with power σ 2 n = σ 2 Â2 . This preprocessing process is similar to some preprocessings present in the RF fingerprinting domain [33].…”

“…Some authors proposed smaller architecture for feature-learning approaches in RF fingerprinting. Arroyo et al presented a low complexity trainable architecture with 1757 trainable parameters [33]. For our part, the model trained in Section 4.2 was composed of 2282 complex parameters.…”