Classification of Network Data with Machine Learning Methods for Intelligent Intrusion Detection Systems

In information systems, it has become very important to store personal and institutional information and access it safely and quickly when necessary. To ensure the confidentiality of information against unauthorized access, institutions or organizations must protect their important data securely and take various precautions. Intrusion detection systems (IDS) are among these measures. One of the issues that should be carefully considered while creating an IDS is the dataset to be used. In terms of IDS, a dataset is the data obtained from network packets or log records that contain attack data and are necessary to identify attack patterns during the training and testing stages of the system. In this article, widely used machine learning techniques (decision tree, K-nearest neighbor, and support vector machine algorithms) are used to increase the performance of IDSs. The studies were tested on the NSL-KDD dataset, one of the most used datasets in evaluating IDSs. As a result of the tests, it was seen that the highest accuracy rate was 99.7%, and the lowest accuracy rate was 98.7%. The obtained results have shown that the proposed machine learning methods can be used with high sensitivity and accuracy to develop smart IDSs.