Network intrusion detection using artificial neural network : a comparative study of the NSL-KDD dataset | |
| Author | Vivek, Billapati |
| Call Number | AIT RSPR no.CS-25-01 |
| Subject(s) | Computer networks--Security measures Computer security Neural networks (Computer science) Machine learning |
| Note | A research study submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Computer Science |
| Publisher | Asian Institute of Technology |
| Abstract | The increasing sophistication and evolving nature of cyber threats present substantial challenges to the security of computer networks, often exceeding the capabilities of traditional Network Intrusion Detection Systems (NIDS), which struggle to adapt to new attack patterns and maintain consistent detection accuracy. This study presents the development of an efficient NIDS framework utilizing an Artificial Neural Network (ANN) classifier, designed to improve the identification and classification of both known and novel intrusion types. The proposed framework was applied to the NSL-KDD dataset, a widely recognized benchmark for intrusion detection research, and its performance was systematically compared to conventional machine learning models, namely Random Forest and K-Nearest Neighbors.The results demonstrate that the ANN model achieved a strong overall accuracy of 0.79 on the unseen KDDTest set and performed consistently across key metrics, with a precision of 0.82, recall of 0.78, and F1-score of 0.77. These scores exceeded those of the Random Forest (F1 score: 0.73) and K-Nearest Neighbors (F1-score: 0.72) models, underscoring the effectiveness of deep learning architectures in addressing the complexity of multi-class intrusion detection tasks. While the ANN exhibited high accuracy in detecting well-represented classes such as Denial of Service (DoS), Probe, and Normal traffic, it faced persistent difficulties in classifying the rare Remote-to-Local (R2L) and User-to-Root (U2R) attack types. This limitation is attributed to the low occurrence of these attacks and their significant overlap with legitimate network activity in the feature space.The findings highlight the potential of ANN-based models to enhance the effectiveness of intrusion detection systems while also emphasizing the need for future research focused on improving the detection of low-frequency, stealthy threats. |
| Year | 2025 |
| Type | Research Study Project Report (RSPR) |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Chantri Polprasert |
| Examination Committee(s) | Attaphongse Taparugssanagorn;Chaklam Silpasuwanchai |
| Scholarship Donor(s) | AIT Scholarships |
| Degree | Research Studies Project Report (M. Eng.) - Asian Institute of Technology, 2025 |