基于自编码网络的入侵检测算法
首发时间:2019-12-27
摘要:由于入侵检测的特殊性,传统的机器学习算法无法获得良好的泛化性能,而复杂的深度学习算法由于数据集的限制,容易出现过拟合现象。针对入侵检测的数据特征,本文提出了一种基于自编码器的集成学习算法,利用自编码器的重建误差来识别异常数据,同时利用dropout 方法和集成学习来降低过拟合的风险。并通过可视化自编码器的编码结果,来解释自编码的作用和学习的结果。在KDD99数据集上的实验表明:基于自编码器的集成学习模型具有良好的泛化性能和较高的训练速率。
For information in English, please click here
Intrusion detection based on autoencoder
Abstract:Due to the particularity of intrusion detection, the traditional machine learning algorithm can not get good generalization performance, while the complex deep learning algorithm is prone to over fitting due to the limitation of data set.In this paper, according to the data feature of intrusion detectionwe use a ensemble learning method based on autoencoder which use the reconstruction error to identify anomaly data and use dropout method and ensemble learning to reduce the risk of over fitting. We also explain the result of autoencoder by visualizing the output of autoencoder. The generalization performance and training speed is confirm by experiment on kdd99 dataset.
Keywords: cyberspace security instrution detection machine learning autoencoder
基金:
引用
No.****
同行评议
共计0人参与
勘误表
基于自编码网络的入侵检测算法
评论
全部评论0/1000