Enregistré dans:
| Auteur principal: | |
|---|---|
| Format: | Recurso digital |
| Langue: | anglais |
| Publié: |
Zenodo
2025
|
| Sujets: | |
| Accès en ligne: | https://doi.org/10.5281/zenodo.15040140 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
Table des matières:
- <p><em><span>The growing complexity of cyber threats has necessitated the development of innovative solutions to protect modern networks. Intrusion Detection Systems (IDS) have long been a cornerstone of cybersecurity, designed to monitor and detect anomalous activities. However, they are often associated with moderate detection accuracy and high false alarm rates, particularly when encountering complex or novel attack patterns. This underscores the need for further experimentation and refinement.</span></em></p> <p><em><span>This thesis explores the application of machine learning algorithms for intrusion detection using the KDD Cup 1999 dataset, aiming to classify network traffic into different categories, distinguishing between normal activity and various types of malicious threats. In addition to evaluating multi-class classification, a binary classification approach was also explored to determine whether simplifying the detection process improves overall reliability.</span></em></p> <p><em><span>These insights emphasize the importance of selecting models based on the specific needs of an IDS deployment. Additionally, this study highlights several challenges encountered, including high false positive rates, imbalanced class distributions, and the necessity for continuous adaptation in the ever-evolving cybersecurity landscape. Addressing these issues is crucial for developing more adaptive and proactive network security strategies, further strengthening the role of machine learning in modern IDS implementations.</span></em></p> <p><em>This research was originally submitted as a thesis at the University of Macedonia and has been refined for open-access publication.</em></p>