传感技术学报2025，Vol.38Issue(10)：1853-1861,9.DOI:10.3969/j.issn.1004-1699.2025.10.017
基于GAN&CNN的物联网环境下入侵检测研究
Study on Intrusion Detection in IoTs Environment Based on GAN&CNN
卢志成 1徐海峰 2潘巨龙1
作者信息
- 1. 中国计量大学信息工程学院,浙江 杭州 310018
- 2. 丽水职业技术学院电子信息学院,浙江 丽水 323000
- 折叠
摘要
Abstract
As technology advances,network intrusion techniques have become increasingly sophisticated,posing significant security chal-lenges to edge devices within the Internet of Things(IoT)ecosystem.To overcome the limitations of traditional intrusion detection models in IoT environments——particularly their suboptimal detection performance and incompatibility with the resource-constrained nature of edge devices,a lightweight model that integrates generative adversarial networks(GANs)and convolutional neural networks(CNNs)is proposed for identifying intrusion behaviors in IoT settings.This innovative approach leverages GANs to address data imbalance issues,while a lightweight CNN based on the cross stage partial(CSP)structure is employed for efficient traffic feature extraction.Traffic data classification is achieved by using a Softmax function.Experimental results on the UNSW-NB15 and CICIDS2018 datasets demonstrate the model's exceptional performance,achieving accuracy rates of 99.64%and 99.65%,precision rates of 99.55%and 99.35%,recall rates of 99.61%and 99.64%,and F1 scores of 99.58%and 99.49%,respectively.Additionally,the model maintains a compact size of only 21 KB to 32 KB.These findings highlight the model's capability to deliver high-precision intrusion detection with minimal computa-tional and storage requirements,making it an ideal solution for the demanding constraints of IoT environments.关键词
入侵检测/物联网/生成对抗网络/卷积神经网络/跨阶段局部结构Key words
intrusion detection/Internet of Things/generative adversarial network/convolutional neural network/cross stage partial structure分类
计算机与自动化引用本文复制引用
卢志成,徐海峰,潘巨龙..基于GAN&CNN的物联网环境下入侵检测研究[J].传感技术学报,2025,38(10):1853-1861,9.
