文章基本信息

标题：MULTI-CLASSIFICATION OF UNSW-NB15 DATASET FOR NETWORK ANOMALY DETECTION SYSTEM
本地全文：下载
作者：MUKRIMAH NAWIR ; AMIZA AMIR ; NAIMAH YAAKOB 等
期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：15
出版社：Journal of Theoretical and Applied
摘要：Problem to classify more than two classes (called as multi-class) for network anomaly detection system using machine learning techniques are very challenging and become a vital factor when the growth of many network attacks might endanger the performances of network system. A tremendous increase in the various number of network threats compromise the network system motivate the network anomaly detection system to be relevant and necessary to be implement using a powerful tool (machine learning approach) for network security issue. In this work, a model of an Online Average One Dependence Estimator (AODE) algorithm for multi-classification of UNSW-NB15 dataset that high in accuracy with a low false alarm rate (FAR) was built to overcome the issues such as the nature of data (complex data that represent into more than two classes), dynamical data in a network system, and frequent update (for streaming data that need a fast processing). The obtained results from the conducted experiment showed that Online AODE more recently detect the Worms class where the percentage of accuracy for classification is 99.93% with small FAR is only 0.001. Moreover, online AODE is an outperformed based on accuracy compare to online Na�ve Bayes (NB) where the classification rate 83.47% and 69.60% respectively for multi-classification the UNSW-NB15 dataset. Since, the given data is a streaming data in a computer network time need to be enumerated to have a fast algorithm for network anomaly detection system before the network system become in a critical condition. Although, the online NB is most fastest for multi-classification yet online AODE give a comparable result based on processing time.
关键词：Multi-classification; Network Anomaly Detection System; Averaged One Dependence (AODE); Machine Learning; UNSW-NB15 Dataset