An Efficient Privacy-preserving Intrusion Detection Scheme for UAV Swarm Networks

TL;DR

This work presents a lightweight, privacy-preserving IDS for UAV swarm networks using federated continuous learning. It introduces a three-component Encoder-Classifier architecture with a swarm-specific input layer, a shared CNN-LSTM encoder, and swarm-tailored classifiers, enabling multiclass intrusion detection across heterogeneous datasets while preserving data privacy. Experimental results on four diverse UAV-related datasets show high accuracies and favorable computational and communication efficiency, outperforming several baselines. The approach’s decentralized design and continual learning capability address data heterogeneity and forgetting, offering practical impact for secure, scalable UAV swarms and potential advancements toward more generalized autonomous security solutions.

Abstract

The rapid proliferation of unmanned aerial vehicles (UAVs) and their applications in diverse domains, such as surveillance, disaster management, agriculture, and defense, have revolutionized modern technology. While the potential benefits of swarm-based UAV networks are growing significantly, they are vulnerable to various security attacks that can jeopardize the overall mission success by degrading their performance, disrupting decision-making, and compromising the trajectory planning process. The Intrusion Detection System (IDS) plays a vital role in identifying potential security attacks to ensure the secure operation of UAV swarm networks. However, conventional IDS primarily focuses on binary classification with resource-intensive neural networks and faces challenges, including latency, privacy breaches, increased performance overhead, and model drift. This research aims to address these challenges by developing a novel lightweight and federated continuous learning-based IDS scheme. Our proposed model facilitates decentralized training across diverse UAV swarms to ensure data heterogeneity and privacy. The performance evaluation of our model demonstrates significant improvements, with classification accuracies of 99.45% on UKM-IDS, 99.99% on UAV-IDS, 96.85% on TLM-UAV dataset, and 98.05% on Cyber-Physical datasets.

Figures (9)

• Figure 1: The model diagram of the proposed Encoder-Classifier framework.
• Figure 2: Federated learning based system architecture
• Figure 3: Confusion Matrix of Four Heterogeneous Swarm
• Figure 4: Accuracy Comparison across different dataset
• Figure 5: System resource utilization during 50 federated rounds.