HSCP: A Two-Stage Spectral Clustering Framework for Resource-Constrained UAV Identification

TL;DR

This work tackles the challenge of deploying RF fingerprinting-based UAV identification on resource-constrained edge devices by introducing HSCP, a two-stage hierarchical spectral clustering pruning framework that jointly reduces network depth and width via CK A-based similarity of layer and channel representations.By first pruning redundant layers and then redundant channels, HSCP achieves extreme compression with minimal loss—and often a gain—in accuracy, while a Mixup-based noise-robust fine-tuning enhances performance in low-SNR wireless environments.Evaluations on the UAV-M100 dataset across ResNet18, MobileNet-V2, and ShuffleNet-V2 demonstrate substantial parameter and FLOPs reductions (e.g., up to ~86% and ~84% respectively for ResNet18) and improved robustness, including under challenging SNR conditions.The approach advances practical deployments of RFFI for UAVs by delivering a universal, architecture-agnostic pruning framework that preserves discriminative RF fingerprints with strong edge-edge performance and real-time inference capabilities.

Abstract

With the rapid development of Unmanned Aerial Vehicles (UAVs) and the increasing complexity of low-altitude security threats, traditional UAV identification methods struggle to extract reliable signal features and meet real-time requirements in complex environments. Recently, deep learning based Radio Frequency Fingerprint Identification (RFFI) approaches have greatly improved recognition accuracy. However, their large model sizes and high computational demands hinder deployment on resource-constrained edge devices. While model pruning offers a general solution for complexity reduction, existing weight, channel, and layer pruning techniques struggle to concurrently optimize compression rate, hardware acceleration, and recognition accuracy. To this end, in this paper, we introduce HSCP, a Hierarchical Spectral Clustering Pruning framework that combines layer pruning with channel pruning to achieve extreme compression, high performance, and efficient inference. In the first stage, HSCP employs spectral clustering guided by Centered Kernel Alignment (CKA) to identify and remove redundant layers. Subsequently, the same strategy is applied to the channel dimension to eliminate a finer redundancy. To ensure robustness, we further employ a noise-robust fine-tuning strategy. Experiments on the UAV-M100 benchmark demonstrate that HSCP outperforms existing channel and layer pruning methods. Specifically, HSCP achieves $86.39\%$ parameter reduction and $84.44\%$ FLOPs reduction on ResNet18 while improving accuracy by $1.49\%$ compared to the unpruned baseline, and maintains superior robustness even in low signal-to-noise ratio environments.

Figures (5)

• Figure 1: Schematic of the data processing pipeline. Raw signals are processed via normalization, noise injection, and STFT transformation, followed by Mixup augmentation.
• Figure 2: Illustration of the two-stage pruning pipeline. The method sequentially performs layer and channel pruning using a unified CKA-based spectral clustering strategy. In both stages, only the leading components from each cluster are retained while redundant ones are removed, resulting the final structurally compressed model.
• Figure 3: Performance evaluation of pruned ResNet18 under varying SNR. The figure displays accuracy variation with SNR ranging from $-5$ dB to $20$ dB for the baseline methods and ours.
• Figure 4: t-SNE visualization of feature representations extracted by different pruning models at an SNR of 0 dB. The subplots correspond to (a) HRank, (b) Sr-init, (c) PSR, and (d) Ours.
• Figure 5: Confusion matrices of the classification results evaluated at SNR = 0 dB. The subplots correspond to (a) HRank, (b) Sr-init, (c) PSR, and (d) Ours.