基于数据包字节长度的线性自回归(Autoregression)和支持向量分类机(SVM)的网络流量预测建模与分析

Modeling and Analysis of Network Traffic Prediction Using Autoregression and Support Vector Machine Based on Byte Length in Data Packets

本文提供了两种网络流量入侵检测的方法和它们的结果对比。这两种方法分别为线性的自回归预测以及非线性的 支持向量机预测。本文将给出使用这两种方法在预测网络攻击的夺效性的详细分析。实验证明用支持向量机模型确实改进了对 于攻击的识别性能，并且其误警率比 AR 模型低了很多。此外，与 SVM 相比较，AR 预测模型的计算复杂度要低。

This paper presents the results of two methods used in predicting network traffic for intrusion detection.These methods are linear prediction using autoregression,and nonlinear prediction using support vector regression.This paper will give details on the effectiveness of detecting attacks relative to each prediction method.It was found that the SVM for regression improved the detectability of the attacks and lowered the false alarm rates compared with AR model.In addition,the AR model is less computationally expensive when compared with SVM regression.