摘要
提出一种基于递归神经网络(RNN)的方法,用于计算机网络主机异常行为预测,通过使用Kaggle上的计算机网络流量数据集,探讨如何利用RNN模型对网络流量时间序列进行分析,并预测网络流量异常。对数据进行归一化处理,以提高RNN模型的训练效果。构建了多个辅助函数,用于数据集特征和目标的创建、模型训练及预测流量。提出了一种基于近似熵(Approximate Entropy, ApEn)的方法,用于评估时间序列的复杂度和不规则性。研究表明,较高的ApEn值指示着时间序列的高度不规则性,反映出潜在的异常行为。为网络安全监测和异常检测提供了一种新的思路,通过结合RNN模型和时间序列熵分析,可有效地对计算机网络中的异常行为进行早期预警和检测。
The study proposes a method based on recurrent neural network(RNN),predicts the abnormal behavior of computer network hosts.By using the computer network traffic data set on Kaggle,the study discusses how to use RNN model to analyze the time series of network traffic,and predicts the abnormal network traffic.The data is normalized to improve the training effect of RNN model.Several auxiliary functions are constructed for the creation of data set features and targets,model training and traffic prediction.A method based on Approximate Entropy(ApEn)is proposed to evaluate the complexity and irregularity of time series.High ApEn values indicate a high degree of irregularity in the time series,reflecting potentially abnormal behavior.This provides a new idea for network security monitoring and anomaly detection.The combination of RNN model and time series entropy analysis can effectively conduct early warning and detection on the abnormal behavior in computer network.
作者
马玉冰
邢磊
Ma Bingyu;Xing Lei(Shandong Huayu University of Technology,Dezhou 253034,China)
出处
《黑龙江科学》
2025年第8期22-25,共4页
Heilongjiang Science
关键词
递归神经网络
流量数据
异常行为预测
近似熵
计算机网络安全
Recurrent neural network
Traffic data
Abnormal behavior prediction
Approximate entropy
Computer network security
