Tools
Tools
Shaik, Abdur Razzaq (2024) Enhancing Efficiency of Machine Learning Techniques with Feature Selection and Hyperparameter Tuning for Intrusion Detection towards Leveraging Cybersecurity. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (994kB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
Cybersecurity enhancement is a continuous process due to increasing cyberattacks of late. Traditional security mechanisms were based on certain heuristics that could detect intrusions based on their detection process. However, with the emergence of Artificial Intelligence (AI) methods such as machine learning (ML) approaches, learning-based models are found efficient due to their ability to learn from labelled data continuously. It is found in the literature that ML models based on supervised learning show deteriorated intrusion detection performance when training samples are not with designed quantity and quality. Therefore, it is important to leverage performance of ML models with certain optimizations. This is the motivation behind this research which is aimed at building an intrusion detection system based on ML models with feature engineering and hyperparameter tuning optimizations. The system is evaluated using CICIDS2017 dataset. Intrusion detection system is implemented using binomial classification and also multi-class classification. In the binomial classification highest accuracy is achieved by RF model with 99.87%. In case of multi-class classification without optimizations, highest accuracy is exhibited by RF with 99.44%. In case of multi-class classification with optimizations, highest accuracy is exhibited by XGBoost with 99.64%.
Actions (login required)
 |
View Item |