With the growing amount of internet and IoT traffic across the network, network anomaly detection system has become a popular and useful strategy to detect anomalies, attacks and intrusions. With machine learning technique, network traffic anomalies can be detected with reasonable prediction accuracy. However, most of the previous work has been focused on detecting anomalies using traditional machine learning environment. Because of ever increasing amount of data and high speed networks, traditional machine learning environment becomes infeasible to cope with the current condition. In this paper, we investigate the feasibility of the applying one of the big data technologies, Apache Spark, to classify different attacks rather than detecting anomalies. We employ traditional machine learning algorithms, namely, Multinomial Logistic Regression, Decision Tree, Random Forest, Multi-layer perception and Naive Bayes using generated dataset of MAWILab gold standard and classify 15 different attack types. In addition, we investigate the efficiency of Apache Spark in terms of accuracy and speed under varied configuration setting of Spark. Our results demonstrate that employing big data technologies adds more benefits to network traffic anomaly detector than traditional machine learning environment in terms of prediction accuracy and execution time.
