Enhanced DDoS Attack Detection with Autoencoder Deep Learning Models

Sujin, Ebin (2024) Enhanced DDoS Attack Detection with Autoencoder Deep Learning Models. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (2MB) | Preview

Abstract

DDoS (Distributed Denial of Service) is a type of cyberattack where multiple compromised systems are used to flood a target server, service, or network with an overwhelming amount of traffic. This study aims at improving detection of DDoS assaults using the UNSW-NB15 dataset, which comprises a versatile representation of network traffic that is derived from modern normal activities and simulated attacks. To tackle class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) was used and overall results showed an impressive distribution for various attack categories. The proposed models were evaluated based on the following criteria: Machine and Deep Learning models including Decision Tree Classifier, Logistic Regression, Long Short Term Memory (LSTM) and Bidirectional LSTM (Bi-LSTM). Notably, the implementation of Autoencoders for feature extraction was a major improvement in the distances used on the model. About all tested models the highest test set accuracy of 86% was obtained with the use of the Bi-LSTM with Autoencoder as it was able to capture longer sequential dependencies in the network traffic data efficiently. The dataset is initially a binary one that is changed to a multi-class to be able to differentiate nine different types of attacks from normal traffic. The findings of this work also establish the possibility of enriching the relational Deep learning architectures to enhance the reliability of DDoS detection systems that can then support the improvement of cybersecurity systems in contemporary networks. Subsequently, the insights call for further study of other Deep Learning approaches so as to counter new threats in network security successfully.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Horn, Christian UNSPECIFIED
Uncontrolled Keywords:	DDoS Attack; UNSW-NB15 Dataset; Machine Learning (ML); Deep Learning (DL); Bidirectional LSTM (Bi-LSTM); Autoencoders
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	05 Sep 2025 11:11
Last Modified:	05 Sep 2025 11:11
URI:	https://norma.ncirl.ie/id/eprint/8820

Actions (login required)

View Item