Benidir, Ahmed, Ould-Slimane, Hakima et Kara, Nadjia.
2025.
« NEMESIS: An enhanced hybrid intrusion detection system leveraging deep Q-learning and random forest ».
In 16th International Conference on Emerging Ubiquitous Systems and Pervasive Networks / 15th International Conference on Current and Future Trends of Information and Communication Technologies in Healthcare (Istanbul, Turkey, Oct. 28-30, 2025)
Coll. « Procedia Computer Science », vol. 272.
pp. 202-209.
Elsevier B.V..
Prévisualisation |
PDF
Kara-N-2025-33107.pdf - Version publiée Licence d'utilisation : Creative Commons CC BY-NC. Télécharger (820kB) | Prévisualisation |
Résumé
Network intrusion detection systems (NIDS) face the growing difficulty posed by increasingly sophisticated and unseen attacks, which represent a dangerous threat due to their ability to exploit vulnerabilities that have not yet been identified. These attacks are inherently difficult to detect with conventional NIDS because such systems typically are built on known threat patterns or signatures, which are absent in unseen scenarios. Consequently, this greatly limits their efficacy in mitigating advanced threats, making networks susceptible to potential security breaches. To address these challenges, recent years have witnessed the emergence of various Reinforcement Learning (RL) approaches aimed at enhancing the automatic detection of network intrusions. These systems are equipped with autonomous agents that acquire the ability to learn independently and make decisions without requiring direct input or knowledge of human experts. In this paper, we propose a network intrusion detection mechanism that integrates a Deep Q Network-based model (DQN) with a supervised machine learning algorithm specifically designed for attack classification. Our model is characterized by meticulous fine-tuning of hyperparameters to optimize the performance of detection. Extensive experimental evaluations that take advantage of the NSL-KDD and CSE-CICIDS2017 datasets demonstrate that our hybrid approach significantly improves detection accuracy across various types of attack and outperforms other existing state-of-the-art solutions designed for similar purposes.
| Type de document: | Compte rendu de conférence |
|---|---|
| Éditeurs: | Éditeurs ORCID Shakshuki, E. NON SPÉCIFIÉ |
| Professeur: | Professeur Kara, Nadjia |
| Affiliation: | Génie logiciel et des technologies de l'information |
| Date de dépôt: | 17 déc. 2025 15:14 |
| Dernière modification: | 10 janv. 2026 18:35 |
| URI: | https://espace2.etsmtl.ca/id/eprint/33107 |
Actions (Authentification requise)
 |
Dernière vérification avant le dépôt |