Network intrusion detection system using Deep learning Approach

B.Sc. Senait Molla Meressa, BSc

Master's thesis

Abstract:

The increasing rate of network attacks is a concerning issue that affects the availability, confidentiality, and integrity of data. To detect malicious cyberattacks, one should require a genuine security system that automates the process of monitoring the events that occur in a network. In this thesis, an anomaly intrusion detection system has been studied and modeled based on a deep learning approach. The KDD Cup '99 dataset is used during the experimental analysis of this thesis work. All attack types within the selected dataset fall under one of the three TCP/IP layer categories. Thus we specifically propose to build three IDS systems, i.e., application layer IDS, transport layer IDS, and Network layer IDS, based on the malware attack types that appear at the corresponding TCP/IP layers. To this end, the KDD cup '99 dataset has been utilized, and data preparation was performed to split the dataset into three categories of the proposed IDS systems. The proposed intrusion detection systems were individually evaluated using different performance evaluation metrics. Extensive experimental results at each layer show that the proposed method has achieved a higher detection rate (DR) and a low false alarm rate (FAR).

Keywords

Intrusion Detection System, Anomaly Detection, KDD cup '99, False Alarm Rate, Detection Rate, Deep Learning, LSTM

Language used: English

Date on which the thesis was submitted / produced: 31. 3. 2021

Thesis defence

• Supervisor: Ing. Tomáš Vokoun

Czech University of Life Sciences Prague

Faculty of Economics and Management

Master programme / field:
Systems Engineering and Informatics / Informatics