| 引用本文: | 李帅,王先培,王泉德,牛胜巍.基于SMDP强化学习的电力信息网络入侵检测研究[J].电力自动化设备,2006,(12):75-78 |
| .Research on intrusion detection based on SMDP reinforcement learning in electric power information network[J].Electric Power Automation Equipment,2006,(12):75-78 |
|
|
| 摘要: |
| 介绍了电力信息网络总体防护体系结构及安全现状,阐述了在电力信息网中常用的防火墙、入侵检测系统(IDS)等防护手段,分析了当前入侵检测方法及难以确定正常与异常的阀值、误报率和漏报率高的不足。提出了基于半马尔可夫决策过程(SMDP)强化学习的IDS模型。论述了强化学习的理论、算法及衡量标准,马尔可夫决策过程,SMDP在电力信息网络中的应用。改进后的SMDP学习算法,使系统的误报率降低、检测率提高。 |
| 关键词: 电力系统,强化学习,半马尔可夫过程,入侵检测 |
| DOI： |
| 分类号:TM73 TP393.08 |
| 基金项目: |
|
| Research on intrusion detection based on SMDP reinforcement learning in electric power information network |
|
LI Shuai WANG Xian-pei WANG Quan-de NIU Sheng-wei
|
| Abstract: |
| The defense architecture and current security conditions of the electric power information network system are introduced and general defensive technologies such as the firewall and the IDS(Intrusion Detection System)are described.Defects of current intrusion detection methods are analyzed:it is difficult to determine thresholds of normality and abnormality;the false-positive rate and the false-negative rate are high.A model of intrusion detection based on the SMDP(Semi-Markov Decision Process)reinforcement learning is proposed.Its principle,algorithm and standard,as well as the Markov decision process and applications of the SMDP in the electric power information network are discussed.The improved SMDP learning algorithm lowers the false-positive rate and the false-negative rate. |
| Key words: electric power systems,reinforcement learning,SMDP,intrusion detection system |
