摘要
为进一步降低样本成本并加快模型收敛速度,提出基于探索和开发的指数加权算法(exponential-weight algorithm for exploration and exploitation,EXP3)和增量微调卷积神经网络(fine-tuning convolutional neural networks,FCNN)的入侵检测系统(EXP3-FCNN)。利用EXP3算法自适应选择最佳主动学习策略,代替单一的主动学习算法,提高样本质量;利用增量微调卷积神经网络提取流量数据更深层次的特征;使用AWID数据集作为实验数据。实验结果表明,该方案在保证模型精确度、召回率等性能指标的基础上,降低了样本成本,提高了模型的收敛效率。
To further reduce sample cost and speed up model convergence,an intrusion detection system(EXP3-FCNN)based on exponential-weight algorithm for exploration and exploitation(EXP3)and fine-tuning convolutional neural networks(fine-tuning convolutional neural networks,FCNN)was proposed.The EXP3 algorithm was used to adaptively select the best active lear-ning strategy instead of a single active learning algorithm to improve sample quality.The incremental fine-tuning convolutional neural network was used to extract the deeper features of the flow data,and the AWID dataset was used as the experimental data.Experimental results show that the proposed scheme not only reduces the sample cost,but also improves the convergence efficiency of the model on the basis of ensuring model accuracy,recall and other performance indicators.
作者
谷朝阳
王亮亮
李晋国
王雪妍
GU Zhao-yang;WANG Liang-liang;LI Jin-guo;WANG Xue-yan(College of Computer Science and Technology,Shanghai University of Electric Power,Shanghai 201306,China)
出处
《计算机工程与设计》
北大核心
2023年第3期699-706,共8页
Computer Engineering and Design
基金
国家自然科学基金项目(61802249、U1936213)。
关键词
入侵检测系统
主动学习
自适应
基于探索和开发的指数加权算法
样本成本
增量微调神经网络
分层
intrusion detection system
active learning
adaptiveness
exponential-weight algorithm for exploration and exploitation
sample cost
incremental fine-tuning of neural networks
hierarchical