Enhancing Intrusion Detection System Using Osprey Optimization Algorithm with Ensemble Learning Model

Networks play crucial roles in the modern world, and cybersecurity has evolved into a vital research field. An Intrusion Detection System (IDS) is a significant cybersecurity system that monitors the status of hardware and software operating within the network. Even after decades of development, current IDSs still encounter problems enhancing recognition accuracy, decreasing the False Alarm Rate (FAR), and identifying unknown attacks. To resolve the above challenges, several researchers have concentrated on emerging IDSs that exploit Machine Learning (ML) approaches. ML techniques automatically learn the crucial differences between normal and abnormal data with maximum accuracy. Moreover, ML approaches have strong generalizability, enabling them to effectively identify unknown or novel attacks, further bolstering their capabilities in cybersecurity. Deep Learning (DL), a subcategory of ML that utilizes Neural Networks (NNs), is gaining attention for its outstanding performance. This study introduces an Enhanced IDS using the Osprey Optimization Algorithm with Ensemble Learning (EIDS-OSOAEL) model. The EIDS-OSOAEL technique mainly focuses on the design of an ensemble classifier that integrates the output from multiple classes. Primarily, the EIDS-OSOAEL technique involves a min-max scalar for scaling the input data into a uniform format. Besides, the Tandem‐Twirl Modified Bacterial Foraging Optimization Algorithm (TT‐MBFOA) approach is employed to achieve the optimal Feature Selection (FS). For intrusion detection, the EIDS-OSOAEL approach undergoes an ensemble of three classifiers, namely AutoEncoder (AE), Feed-Forward NN (FFNN), and Elman NN (ENN). The OS-OA approach is utilized to adjust the hyperparameter values of these models. The experimental results of the EIDS-OSOAEL technique are evaluated under a standard dataset. The performance validation of the EIDS-OSOAEL technique showed a superior 99.76% over recent approaches.