A Simple and Robust Approach of Random Forest for Intrusion Detection System in Cyber Security

Abstract

An Intrusion Detection System (IDS) is monitor the internet user’s behavior in to the categories of normal or abnormal and record all these activities. It secure the physical or virtual system from threads and viruses. Detection of vulnerabilities in computer network is very complex task. The information of government and private organizations may be leaked or damaged by attackers in form of viruses. The information security is very important aspect to protect the valuable data of any organization.In the presented paper, a robust model proposed for IDS that uses a random selection algorithm for feature selection and a random forest method using a random forest classifier for designing a highly robust IDS model. In the machine learning area, Random Forest (RF) is a classifier method that is most effective compared with other classifier methods under attack classification. To make the proposed model more robust, we design a hybrid model by performing two experiments. The random forest classifier is used in the first experiment, whereas we have developed a random feature selection method for test. For experimenting on IDS, a restructured form of the KDD-99 dataset is used named as NSL-KDD data set and measure the performance in terms of FAR (False Alarm Rate), high accuracy, high DR (Detection Rate) and Mathews Correction Coefficient (MCC) to check the robustness of the model.