Malware Classification Using Deep Learning
Lokesh J, Ahemad Talwar, Danesh H M, Danthuluri Sudha, NAGENDRAN NIVETHA, Junaid Mundichipparakkal · 2025
The rapid advancement of malware implies that malware detection and malware detection methods must advance as well. Signature detection methods are no longer able to adapt to newly evolving malware threats, hence there is a real need for novel detection techniques. In this regard, (ML) and (DL) detection methods serve as effective alternatives. This research studied various ML and DL methods to classify malware using both malicious and benign datasets. The evaluation of different methods was based on accuracy, recall, and precision. It was found that deep learning techniques offer a promising avenue for cost-effective solutions to classify files from both known malware and benign samples in the field of cybersecurity. To this end, we evaluated different methods including Random Forest, Support Vector Machines (SVM), and Gradient Boosting, in comparison to deep learning approaches like Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). Additionally, we proposed various methods including combinations of static and dynamic techniques, which improved detection rates against previously unknown malware strains. Both deep learning models showed promising results. Notably, RNN performance achieved a true positive rate accuracy of 99.11% outperforming conventional ML models while CNN reached an accuracy of 96.7%. These results demonstrate that models based on RNNs are well-suited for stream-based classification of malware samples. Finally, this research contributes to the body of knowledge by proposing an optimized framework capable of combating real-time active malware threats. For ongoing research, the framework can be extended to include better explainable AI techniques to enhance transparency for both laypersons and malware classification models.