A comparative analysis of different classification techniques for cloud intrusion detection systems' alerts and fuzzy classifiers

Intrusion detection system (IDS) as one of huge research problem in network security is the most effective tool of protection. It is a method of parsing network traffic data to detect security abuses. Data mining can play a very significant role in evolving an IDS. The dataset of IDSs or soft computing techniques based IDS can be classified into normal and abnormal traffic in order for generated alerts to detect threats. In this paper, we utilised the most common classification algorithms: Decision Tree (J48), Naive Bayes, OneR, and K-Nearest Neighbour (K-NN). These algorithms were chosen after investigating the most effective classification algorithms that are widely used. The aim of this study is to present a comparative study for the performance of each system that was gained from our previous experiments: SnortIDS, SuricataIDS, FL-SnortIDS, and FL-SuricataIDS in order to test which classifier algorithm is the best for our systems results, and investigate which system presents significant results. The performance of these classification algorithms was evaluated using 10-fold cross validation. Experiments and assessments of these methods were performed in the WEKA environment using the ISCX dataset.