基于机器学习的传感网核心节点漏洞检测仿真
Simulation of Vulnerability Detection of Sensor Network Core Nodes Based on Machine Learning
徐寅森 1李红艳 1张子栋2
作者信息
- 1. 商丘学院计算机工程学院,河南 商丘 476000
- 2. 集美大学计算机工程学院,福建 厦门 361021
- 折叠
摘要
传感网的核心节点具有能量受限、难补给的特点,导致节点轮休时易出现的覆盖漏洞问题,造成传感网监测盲区.为此提出基于机器学习的传感网核心节点漏洞检测方法.利用支持向量机树形多分类器获取核心节点的位置.采取主成分分析法提取核心节点特征,将其输入到LSTM长短记忆神经网络模型中,并利用滑动窗口与哈希函数训练漏洞检测分类模型,完成传感网核心节点的漏洞检测.实验结果表明,研究方法检测传感网漏洞时平均耗时为 13.6ms,检测率和准确率均可高达 95%,计算得到性能消耗低于10%,90%的用户响应时间均在 50ms以内.
Abstract
At present,the core node in the sensor network has the characteristics of limited energy and difficult supply,leading to the blind area of sensor network monitoring.Therefore,a method of detecting the vulnerability of core nodes in the sensor network was proposed based on machine learning.At first,the tree-based support vector ma-chine multi-classifier was used to obtain the location of the core node.Then,the principal component analysis method was used to extract the characteristics of core nodes and input them into the LSTM long-short memory neural network model.Meanwhile,the sliding window and hash function were used to train the vulnerability detection classification model.Finally,the vulnerability detection of core nodes in the sensor network was completed.Experimental results prove that the average time of detecting sensor network vulnerabilities is 13.6ms.The detection rate and accuracy can reach 95%,and the performance cost is less than 10%.In addition,the response time of 90%of users is within 50ms.
关键词
支持向量机树型多分类器/特征提取/主成分分析/线性哈希函数/欧氏距离Key words
Support vector machine tree multi-classifier/Feature extraction/Principal component analysis/Linear Hash function/Euclidean distance引用本文复制引用
基金项目
河南省高等学校精品在线开放课程建设项目(教高[2019]671)
出版年
2024
NETL
NSTL
万方数据