Enhancing Cyber Security Via Malware Classification using Tuna Swarm-Based Feature Selection with Optimal Deep Learning

Abstract

Malware detection is a central aspect of cyber security that includes detecting and mitigating malicious software, often called malware, that can compromise the safety and integrity of a computer system. Different malware detection techniques are used, such as Machine Learning (ML), signature-based detection, and behavioural analysis. Cutting-edge ML models are widely applied for malware detection. These techniques analyze large datasets to detect features and patterns related to malicious behaviours. Supervised learning trains models on labelled datasets, while unsupervised learning can identify anomalies in system behaviours without predefined labels. Deep Learning (DL)-based malware detection improves the capability to identify polymorphic and sophisticated risks and promotes a more adaptive proactive cyber security system. This study introduces Malware Recognition and Classification using the Tuna Swarm Optimization-based Feature Selection with DL (MRC-TSOFSDL) approach. In the MRC-TSOFSDL approach, the feature subset selection process is accomplished using the TSO model. The Stacked Sparse Autoencoder (SSAE) method is used to recognise malware automatically. Chimp optimization Algorithm (ChoA) based on a hyperparameter tuning process is utilized to improve the malware detection outcomes of the SSAE model. The performance analysis of the MRC-TSOFSDL method is examined under a malware dataset. The comparative results of the MRC-TSOFSDL technique demonstrated a maximum accuracy value of 98.65% over existing models.