«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

29

期数:

2019年01期

页码:

118-123

栏目:

安全与防范

出版日期:

2019-01-10

文章信息/Info

Title:

Design of Distributed Intelligent Intrusion Prevention Scheme Based on Reinforcement Learning

文章编号:

1673-629X(2019)01-0118-06

作者:

李炳星;  季薇

南京邮电大学 通信与信息工程学院,江苏 南京,210003

Author(s):

LI Bing-xing; JI Wei

School of Telecommunications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China

关键词:

协作频谱感知;  入侵防御;  强化学习;  信誉模型

Keywords:

cooperative spectrum sensing; intrusion prevention; reinforcement learning; reputation model

分类号:

TP181

DOI:

10. 3969 / j. issn. 1673-629X. 2019. 01. 025

文献标志码:

A

摘要:

协作频谱感知技术能够有效地提高频谱利用率,然而恶意用户的存在将极大地影响整个认知网络的性能,因此恶意用户的检测与防御变得尤为重要.为有效抵御协作频谱感知中的恶意用户及其恶意行为,提出一种强化学习与信誉模型相结合的分布式智能入侵防御方案.在该方案中,认知用户通过不断地观测和学习周围的环境,对于实时变化的网络环境做出最优的选择,选取最优的邻居用户进行合作,以获得最大的回报.当融合中心的判决结果与认知用户的判决结果一致/不一致时,给予相应信誉值的奖励/惩罚,当信誉值低于一定判决门限视为潜在的恶意用户.最终使得智能的恶意用户主动放弃恶意攻击,开始发送正确的感知值,达到一致性融合.仿真结果表明,所提方案能够有效地抵御恶意攻击,极大地提高了网络的健壮性与稳定性.

Abstract:

Cooperative spectrum sensing technology can effectively improve the utilization of spectrum,but the existence of malicious users will greatly affect the performance of the whole cognitive network. Therefore,the detection and defense of malicious users is particularly important. In order to effectively resist malicious users and malicious behaviors in cooperative spectrum sensing,we propose a dis- tributed intelligent intrusion prevention scheme combining reinforcement learning with reputation model. In this scheme,the cognitive us- er can make the best choice for the real-time changing network environment by constantly observing and learning the surrounding envi- ronment,and select the best neighbor users to cooperate to get the maximum reward. When the result of the fusion center is consistent/ inconsistent with the result of the cognitive user,the reward/ punishment for the corresponding reputation value will be given,and it is considered a potential malicious user when the reputation value is less than a certain decision threshold. Finally,it makes the malicious users of intelligence give up malicious attacks and begin to send the correct sensing values to achieve consensus. The simulation shows that the proposed scheme can effectively resist malicious attacks and greatly improve the robustness and stability of the network.

常用功能

更新日期/Last Update: 2019-01-10