Back to search

University of the Aegean

Advanced machine learning methods for network intrusion detection

Abstract

dc:description

The contemporary Information and Communications Technology infrastructures have become undoubtedly the ``land of opportunity'' for ill-motivated entities, which aim to threaten the confidentiality, integrity and availability of the underlying systems. The ever-increasing magnitude and sophistication of cyber attacks leave no room for rest to the defenders. In this context, the quest for full-fledged and versatile defensive frameworks and methodologies is of high priority. In this direction, Intrusion Detection and Response Systems are essential entities in a network topology aiming to safeguard the protected systems and provide remediation actions against offensive incidents. However, such mechanisms need to be supported by intelligent methods to sustain a high operational capability. In this context, this doctoral thesis focuses on advanced machine learning methods that can deliver beneficial characteristics to intrusion detection and response systems.More specifically, this Phd thesis comprises three tightly interrelated axes, namely a) the provisioning of optimal countermeasures in the context of intrusion response systems, b) the induction of accurate detection rules to enable misuse network intrusion detection, and c) the integration of self-adaptation properties to those systems.In relation to the first axis, this work provides a comprehensive analysis on reactions frameworks which aim to provide cost-benefit countermeasures against cyber attacks. Our analysis aims to critically scrutinize the pertinent works in this field, to pinpoint the Artificial Intelligence methods utilized by them, and to offer an in-depth discussion and side-by-side comparison among them based on several criteria. Also, an extensive discussion is offered to highlight on the shortcomings and future research challenges and directions in this timely field of research.Driven by the fact that reactive frameworks should be triggered upon accurate predictions on the nature of offensive incidents, the second axis of the doctoral thesis at hand focuses on the design and implementation of a rule induction methodology, called Dendron, for misuse intrusion detection systems. More specifically, our methodology takes advantage of both Decision Trees and Genetic Algorithms for the sake of evolving linguistically interpretable and accurate detection rules. Dendron is able to rightly designate the category where attacks belong to, and achieves superior results over other legacy techniques under several classification metrics. Additionally, with the aim of tackling the major limitation of misuse intrusion detection systems to adapt to new network conditions, the third axis pursues the development of a self-adaptive methodology, which can revitalize a detection engine through the automation of its retraining process. Considering the extended size of modern networks and the complexity of big network traffic data, the adaptation problem exceeds the limits of human managing capabilities. Thus, through the utilization of deep-learning based methods, our approach is able to grasp an attack's nature based on generalized feature reconstructions stemming directly from the unknown environment and its unlabeled data. The experimental results reveal that our methodology can breathe new life into an intrusion detection system, thus outperforming rigid detection approaches.

Degree

thesis:*
Grantor dc:publisher
University of the Aegean
Year dc:date
2019

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Papamartzivanos, Dimitrios
  • Παπαμαρτζιβάνος, Δημήτριος

Subjects

dc:subject × 14

Rights

dc:rights
Statement dc:rights
  • BY_NC_ND
Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/45893

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Papamartzivanos, Dimitrios; Παπαμαρτζιβάνος, Δημήτριος. Advanced machine learning methods for network intrusion detection. University of the Aegean, 2019. http://hdl.handle.net/10442/hedi/45893