ZHOU Yanjun 1DUAN Zhangjue1
作者信息
- 1. The Fifth Research Institute of telecommunications technology,Chengdu 610000,China
- 折叠
摘要
Abstract
With the widespread deployment of Unmanned Aerial Vehicles(UAVs)in logistics,surveillance,and communication relay appli-cations,network security challenges have become increasingly prominent.The open and distributed nature of UAV networks makes them highly vulnerable to various cyber attacks,while traditional centralized intrusion detection systems face privacy risks and scalability limitations when processing distributed data.To address this challenge,this paper proposes a distributed network intrusion detection framework based on Federat-ed Learning and Long Short-Term Memory networks(FL-LSTM).The framework enables collaborative training across multiple UAV nodes through federated learning mechanisms,where each node only uploads model parameters without sharing raw data,effectively preserving data pri-vacy.Meanwhile,leveraging the powerful temporal modeling capability of LSTM networks,the system can accurately capture time dependencies in network traffic,improving anomaly detection accuracy.Experiments were conducted using the CIC-IDS2017 dataset,and results show that af-ter 3 rounds of federated training,the system's detection accuracy improved from 83.18%to 84.05%,validating the effectiveness of the proposed method.This research provides a new technical pathway for building secure and efficient UAV network protection systems.关键词
联邦学习/长短期记忆网络/无人机/网络入侵检测/隐私保护/分布式学习Key words
Federated learning/Long short-term memory network/Unmanned aerial vehicle/Network intrusion detection/Privacy protec-tion/Distributed learning分类
信息技术与安全科学
