Building a new artificial intelligence based security model for intrusion detection and prevention systems

With today's technological developments, both the variety and the number of cyber-attacks are constantly increasing. One of the most widely used technologies to prevent cyber-attacks is Intrusion Detection Systems (IDS). In this thesis, studies have been carried out to improve the performance of IDS systems. After examining the related studies in the literature in the first section, STS datasets that are frequently used in the literature are discussed in the second section. Considered datasets are classified by proposed machine learning methods. In the third section, the Comprehensive Cyber Security Intrusion Detection Dataset (CCiDD) dataset was created in order to eliminate the deficiencies of the STS datasets in the literature. The performance of the classifiers has been increased by applying feature selection and hyperparameter optimization processes for all datasets used in this section. In the fourth section, a new dataset called Switch port Anomaly based Intrusion Detection System (SPA-IDS) was created by considering the anomaly increases in switching devices. The created dataset was classified with the recommended machine learning methods after going through the proposed vertical mode decomposition (VMD) and iterative neighborhood component analysis (INCA) steps. In the fifth section, a dataset called Internet of Behavior (IoBe) was created in order to analyze the internet behavior of users. The best features of the created IoBe dataset were selected with the ReliefF feature selector after four pooling operations and then classified with the Bagged Tree algorithm. In the last and sixth section, the concept of Internet of Medical Things (IoMT) is discussed and a new model consisting of feature selection and hyperparameter optimization steps is proposed to detect attacks against these objects. In this thesis, intrusion detection systems are looked at from a broader perspective and new approaches and datasets for the development of STS systems are proposed. In addition, a new dataset was created to analyze the internet behavior of users. Thus, a contribution to the literature has been made with the created datasets and the suggested approaches.