Malware Detection: A Comparison of Different Machine Learning and Deep Learning Networks
K Abirind, K Vijai, R Nandakumar · 2024
Malware poses a significant threat in the modern world as it can maliciously affect users’ data. To combat this menace various machine learning techniques have emerged as potent tools for detecting and mitigating malware attacks. In a bid to discern malware from benign software several machine learning and deep learning models were compared. The dataset sourced from GitHub comprised samples of both malware and non-malware instances. Through rigorous training and testing on this dataset the models’ accuracies were evaluated. Results revealed that the Multilayer Perceptron (MLP) model achieved the lowest accuracy clocking in at 93%. Conversely the ExtraTree Classifier model outperformed others boasting an impressive accuracy of $\mathbf{9 9. 5 4 \%}$. Other models tested included Random Forest with 98.38% accuracy, Decision Tree with $\mathbf{9 8. 8 \%}$, Recurrent Neural Network (RNN) with 96.4% Convolutional Neural Network (CNN) with 96.6% and KNearest Neighbors (KNN) with $\mathbf{9 8. 8 2 \%}$ accuracy. These findings highlight the effectiveness of the ExtraTree classifier in accurately identifying malware.