HYBRID ARTIFICIAL ECOSYSTEM-BASED OPTIMIZATION WITH LIGHT GRADIENT BOOSTING FOR INTRUSION DETECTION
Mohammed Eltahir Abdelhag1, Saad Mamoun 2, Mohd Sarfaraz1, Emad Addin Alsheikh3,1Department of Information Technology and Security, Jazan University, Jazan 45142, Saudi Arabia
2Imam Mohammad Ibn Saud Islamic University (IMSIU), College of Shari’a & Islamic Studies, AL-Ahsaa. Department Computer Science and Information. Saudi Arabia
3 Department of Computer Science, Jazan University, Jazan 45142, Saudi Arabia
mohedtahir@gmail.com1, smaahmed@imamu.edu.sa2, msarfaraz@jazanu.edu.sa1, ialsheikh@jazanu.edu.sa3
Abstract
The number of people and applications make internet usage dramatically increased by users over the past several years, resulting in more security problems. To provide a secure environment, businesses and institutions give more attention to providing more effective safeguards against modern attacks. Machine Learning (ML) algorithms show great positional to be used in Intrusion Detection (ID) systems which can monitor and distinguish whether a packet is a malicious or typical system behavior based on the data it contains. Therefore, this paper introduces an efficient model based on Artificial Ecosystem-based Optimization (AEO) and Light Gradient Boosting model (LGBM), named AEO-LGBM. The AEO is employed to select the most informative features from large datasets, while the LGBM model is used for classification. The AEO-LGBM is verified on two datasets for ID: KDD CUP99 and NSL-KDD and compared to Logistic Regression (LR), Multilayer Perceptron (MLP), and Support Vector Machine (SVM). The results validate the superiority of the introduced model over the examined techniques and recently reported models in the literature for ID.
Keywords: Feature selection, machine learning, Intrusion Detection, metaheuristic algorithms,