Détails Publication
Intrusions Detection and Classification Using Deep Learning Approach,
Discipline: Informatique et sciences de l'information
Auteur(s): Léonard M Sawadogo, Didier Bassolé, Gouayon Koala, Oumarou Sié
Auteur(s) tagués: BASSOLE Didier ; SIE Oumarou
Renseignée par : BASSOLE Didier
Résumé

In this paper we propose an intrusions detection technique using Deep Learning approach that can classify different types of attacks based on user behavior and not on attacks signatures. The Deep Learning approach used is Supervised Learning model called Convolutional Neural Networks (CNN) coupled with Tree Structure whose set is named Tree-CNN. This structure allows for incremental learning. This makes the model capable of learning how to detect and classify new types of attacks as new data arrives. The model was implemented with TensorFlow and trained with the CSE-CIC-IDS2018 dataset. We evaluated the performance of our proposed model and we made comparisons with other approaches considered in related works. The experimental results show that the model can detect and classify intrusions with a score of 99.94% for the detection and 97.54% for the classification.

Mots-clés

930
Enseignants
5632
Publications
49
Laboratoires
84
Projets