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

2025-12-05

AI Summary

Radio-frequency fingerprint identification (RFFI) for unmanned aerial vehicles (UAVs) on resource-constrained edge devices faces a fundamental trade-off among model lightness, recognition accuracy, and noise robustness. Method: This paper proposes a hierarchical spectral clustering pruning framework, featuring the first integration of centered kernel alignment (CKA)-guided hierarchical and channel-level structured pruning, coupled with a noise-robust fine-tuning strategy. Results: On the UAV-M100 dataset, our method compresses ResNet18 by 86.39% in parameters and reduces FLOPs by 84.44%, while improving classification accuracy by 1.49% over the baseline—outperforming existing pruning approaches. Crucially, it maintains superior robustness under low signal-to-noise ratio (SNR) conditions. To the best of our knowledge, this is the first work achieving simultaneous optimization of compression ratio, inference acceleration, and identification performance, thereby enabling practical deployment of edge-based RFFI for low-altitude security applications.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Security and Privacy: Security and privacy of machine learning and AI applications

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.

Problem

• Optimizing UAV identification for resource-constrained edge devices
• Reducing model size and computational demands of deep learning RFFI
• Balancing compression rate, hardware acceleration, and recognition accuracy

Innovation

• Hierarchical spectral clustering pruning framework
• Two-stage redundancy removal via spectral clustering
• Noise-robust fine-tuning for enhanced robustness