Advanced Malicious Behavior Classification Using a Refined ANN-CNN Model

A Hybrid Deep Learning Approach for Enhanced Cybersecurity Threat Detection

Abstract

The rapid growth of the internet has led to an overwhelming increase in online data. Activities such as data transfer, online banking, and business transactions are now conducted over the internet, which, while providing convenience, also presents opportunities for malware developers to exploit vulnerabilities. Cybercriminals use sophisticated methods to bypass security measures, stealing personal data and demanding ransom from victims. To address these growing threats, there is a critical need for more advanced AI-based methods to detect and prevent malware attacks.

In this paper, we propose an improved hybrid ANN-CNN sequential model designed to enhance malware classification performance. Class imbalance is addressed using the SMOTE technique, which ensures that all classes are equally represented. Additionally, Principal Component Analysis (PCA) is employed for feature selection, enabling the model to focus on the most meaningful features and improving both training efficiency and model accuracy.

The model is evaluated on three multiclass datasets: WSN (Wireless Sensor Network), Microsoft Malware, and Virus Malware Digit. The proposed model achieved 98.1%, 99.6%, and 99.0% accuracy, respectively, demonstrating its effectiveness in handling complex, imbalanced, and diverse malware datasets.