Golden jackal optimization algorithm with deep learning assisted intrusion detection system for network security

Abstract

Network security is essential to our daily communications and networks. Cybersecurity researchers initiate the significance of emerging proficient network intrusion detection systems (IDS) for providing secure networks. While attackers endure to progress novel kinds of attacks and the network sizes endure to develop, the necessity for effectual IDS becomes significant. Additionally, IDS aims to offer confidentiality, integrity, and availability for the data communicated in networked computers by avoiding illegal access to the network. Several studies executed machine learning (ML) systems for emerging effectual IDS; then, with the advent of deep learning (DL) systems and artificial neural networks (ANNs) that create features automatically without human interference, researchers started to depend on DL approaches. This study introduces a new Golden Jackal Optimization Algorithm with Deep Learning Assisted Intrusion Detection System for Network Security (GJOADL-IDSNS) technique. The major intention of the GJOADL-IDSNS system lies in the effectual recognition and classification of the intrusions, to achieve network security. Primarily, data normalization is performed to scale the input data into a useful format. In the presented GJOADL-IDSNS technique, the GJOA-based feature selection (GJOA-FS) approach can be employed to elect an optimum subset of features. Next, the GJOADL-IDSNS methodology applies the attention-based bi-directional long short-term memory (A-BiLSTM) model. For hyperparameter tuning of the A-BiLSTM model, the GJOADL-IDSNS technique uses the salp swarm algorithm (SSA). The simulation value of the GJOADL-IDSNS technique has been tested utilizing benchmark datasets. The comparative results stated that the GJOADL-IDSNS technique achieves better performance than other models.