Work place: Department Of Computer Science, American International University-Bangladesh, Dhaka, Bangladesh
Research Interests: Applied computer science, Computational Science and Engineering, Computer systems and computational processes, Computer Architecture and Organization, Theoretical Computer Science
Faria Farzana Dola is born in Bangladesh on 24th September 1999. Currently she is pursuing her Bachelor’s degree in Computer Science and Engineering under the Faculty of Science and Technology at American International University- Bangladesh.
DOI: https://doi.org/10.5815/ijieeb.2022.02.01, Pub. Date: 8 Apr. 2022
Network intrusion is an increasing major concern as we are rapidly advancing in technology. To detect network intrusion, Intrusion Detection Systems are required. Among the wide range of intrusion detection technologies, machine learning methods are the most appropriate. In this paper we investigated different machine learning techniques using NSL-KDD dataset, with steps of building a model. We used Decision Tree, Support Vector Machine, Random Forest, Naïve Bayes, Neural network, adaBoost machine leaning algorithms. At step one, one-hot-encoding is applied to convert categorical to numeric features. At step two, different feature scaling techniques, including normalization and standardization, are applied on these six selected machine learning algorithms with the encoded dataset. Further in this step, for each of the six machine learning algorithms, the better scaling technique application outcome is selected for the comparison in the next step. We considered six pairs of better scaling technique with each machine learning algorithm. Among these six scaling-machine learning pairs, one pair (Naïve Bayes) is dropped for having inferior performance. Hence, the outcome of this step is five scaling-machine learning pairs. At step three, different feature reduction techniques, including low variance filter, high correlation filter, Random Forest, Incremental PCA, are applied to the five scaling-machine learning pairs from step two. Further in this step, for each of the five scaling-machine learning pairs, the better feature reduction technique application outcome is selected for the comparison in the next step. The outcome of this step is five feature reduced scaling-machine learning pairs. At step four, different sampling techniques, including SMOTE, Borderline-SMOTE, ADASYN are applied to the five feature reduced scaling-machine learning pairs. The outcome of this step is five over sampled, feature reduced scaling-machine learning pairs. This outcome is then finally compared to find the best pairs to be used for intrusion detection system.[...] Read more.
Subscribe to receive issue release notifications and newsletters from MECS Press journals