Evaluating and Comparing the Efficiency of Malware Identification for Cyber Security Using Machine Learning Algorithms

Abstract

Malware detection is a critical task in information security, and it should be able to detect the occurrence of potential malicious code in memory. Detection methods typical of the past could sometimes not detect the new malware variants because it is highly inaccurate. In this study, it presents a robust malware detection model employing Long Short-Term Memory (LSTM) networks to deal with class imbalance and feature redundancy. Within a well-structured preprocessing pipeline, it uses feature selection through correlation analysis and class balancing with SMOTE. Experimental results demonstrate that the model performs quite accurately, has an F1 score of 97.97%, accuracy of 99.17%, precision of 98.80%, and recall of 99.17%, proving its reliability and robustness when classifying both benign and Malware instances. Comparative analysis with traditional models, such as Decision Tree and the suggested LSTM model, beats these techniques with the maximum accuracy of 99.09%, according to K-Nearest Neighbors (KNN). The results indicate that the model outperforms malware detection, has good generalization ability and low overfitting, and therefore is a very good solution for real-world cybersecurity applications.