Patrik is a research assistant specialized in computer networking and information systems security. In KInIT, he is a part of the Security team, about to work on anomaly detection and other security-related projects. His other interests include penetration testing, digital forensics, and parallel algorithms.
He is a fresh graduate of the Faculty of Information Technology, Brno University of Technology, from August 2021. During his studies, he was primarily focused on DDoS attacks detection and mitigation as an R&D engineer in CESNET z.s.p.o. Both his Bachelor’s and Master’s theses discussed DDoS mitigation – the first using traditional and the latter utilizing machine learning techniques. In CESNET, he designed, implemented, and experimentally evaluated various algorithms integrated into CESNET’s DDoS Protector as a part of the VI20192022137 Adaptive Protection Against DDoS Attacks project.
He is currently a PhD student at KInIT, doing his PhD at the Faculty Of Information Technology, Brno University of Technology.
PhD topic: Network Intrusion Detection Using Machine Learning
Supervising team: Daniela Chudá (KInIT)
Nowadays, many types of cyberattacks aiming to compromise information systems exist in-the-wild. In order to protect information and communication infrastructure, these attacks must be detected and prevented from succeeding. Such goals are typically achieved by Intrusion Detection Systems (IDSs), which monitor the target system or a network and alert on potential malicious or anomalous behavior. Our research focuses on the domain of network intrusion detection using machine learning (ML). ML utilization has greatly improved attack detection capabilities and made the systems significantly more flexible. However, machine learning is heavily dependent on the data, which is one of the main issues within the domain currently. In order to address the data problem, our research aims to pursue the three areas of interest built around the data-centric artificial intelligence approach. Research efforts will be focused on data augmentation, feature engineering, and transfer learning. By achieving these preset research goals, we believe that ML-based IDSs will become more robust by training on higher-quality data and, thus, be more viable for real-world usage.