International Journal of Applied Information Systems |
Foundation of Computer Science (FCS), NY, USA |
Volume 12 - Number 39 |
Year of Publication: 2022 |
Authors: Bukola Fatimah Balogun, Kazeem Alagbe Gbolagade, Ayisat Wuraola Asaju- Gbolagade |
10.5120/ijais2022451929 |
Bukola Fatimah Balogun, Kazeem Alagbe Gbolagade, Ayisat Wuraola Asaju- Gbolagade . Feature Selection based on Bat Algorithm and Residue Number System for Intrusion Detection System. International Journal of Applied Information Systems. 12, 39 ( June 2022), 32-37. DOI=10.5120/ijais2022451929
The Internet has grown rapidly in the last ten years. Consequently, the interconnection of computers and network devices has become so complex for monitoring that even the security experts do not fully understand its deepest inner workings. Personal computers have become very fast every year. It is not rare for a very ordinary person to connect to the Internet through 20 Mbs lines or faster. With this huge network data, the network security has become very important for monitoring the data. Machine Learning (ML) is a variant of Artificial Intelligence (AI) which uses algorithms to train and make accurate predictions on data. Dimensionality reduction in ML is used to remove redundant or irrelevant features, thereby improving the performance of classification. Chinese Remainder Theorem (CRT) is a modular arithmetic often used as a backward conversion algorithm in Residue Number System (RNS) to solve simultaneous linear congruence using a set of pair-wise relatively prime integers known as moduli set. In this paper, the Intrusion Detection System model was presented. The hybridized Bat algorithm and Chinese Remainder Theorem was used for feature selection and subsequently feature extraction was performed with Principal Component Analysis (PCA). The classification was done utilizing Naïve Bayes (NB). The dataset used for the experimental analysis was the Network Security Laboratory Knowledge Discovery Dataset (NSLKDD). In the experimental phase, 75% of the dataset was used for training and 25% for testing. The results obtained were measured in terms of accuracy, recall, sensitivity, specificity and precision.