| Abstrakt: |
In recent decades, the Internet of Things (IoTs) based network intrusion detection (ID) remains a challenging research topic. Currently, several machine-learning methodologies are extensively used for network ID. Most of the existing methodologies failed to obtain consistent performance in multiple class classification. In this research article, a new ID system is implemented for detecting network intrusions more efficiently. After acquiring the data from UNSW-NB15 and NSL-KDD datasets, the data denoising techniques like min–max scalar and adaptive synthetic sampling are utilized to address the data imbalancing problems. Then, the Firefly Optimization Algorithm (FOA) is implemented to choose the optimal attributes for better Intrusion attack classification. In the final phase, the selected attributes are given as input to the ensemble classifier to classify the normal and attack labels. In this article, the ensemble classifier has four classifiers like K-nearest neighbors, support vector machine, long short term memory and the multi-layer perceptron's. The experimental examination states that the FOA based ensemble model achieved 98.89% and 98.41% of detection rate on the UNSW-NB15 and NSL KDD. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink