International Journal of Computing and Digital Systems | 2026
Authors: Nabat Z.M.; Shnain S.A.; Wotaifi T.A.; Matloob A.Z.K.
DOI: 10.12785/ijcds/1571128789
Journal: International Journal of Computing and Digital Systems
Year: 2026
Publisher: University of Bahrain
Document Type: Article
Open Access: All Open Access; Gold Open Access
Cited by: 0
Intrusion detection systems (IDS) are important in determining malicious and dynamic network traffic patterns that quickly change according to the characteristics of the network. Recent developments have been characterized with most IDS being improved by using machine learning algorithms hence, enhancing their ability to identify cyber threats. However, IDS often have a high rate of false positives, and they create the wrong and ill-timed alerts whenever an anomalous activity is identified. False-alarm rate is significantly high and it reduces the efficiency of the IDS making it harder to identify the actual threats and the actual traffic anomaly. This therefore adds a complex element in controlling cybersecurity risks and it is costly and operative in terms of time. The current research examines a new nature-based evolutionary algorithm (EA), which is the Buzzards Optimization Algorithm (BUZOA), and its combination with artificial neural networks (ANN) to supplement the intrusion detection abilities. An enhanced ensemble neural network (ENN) framework is developed by integrating EA and ANN. In this method, a feed-forward multi-layered artificial neural network (BUZOA-ANN) is utilized to identify emergent, unknown cyber threats with high accuracy through the use of BUZOA. Specifically, the process of global optimization of ANN weights and biases of BUZOA helps to suppress local minima convergence and to reduce the problem of false alarm that plagues most forms of training based on the use of gradients. The experiments of the NSL-KDD dataset gave a high level of accuracy of 96.775 percent and the model proposed is able to perform a better classification than previous research works which indicates that the model is more efficient and effective towards more dependable intrusion detection systems. © 2026, University of Bahrain. All rights reserved.
Artificial neural network; Denial of service; evolutionary algorithm; Intrusion detection system