Multi-Class Malware Detection using modified GNN and Explainable AI
Premanand Pralhad Ghadekar, Tejas Adsare, Neeraj Agrawal, Dhananjay Deore, Tejas Dharmik · 2024
In recent years, the rapid proliferation of sophisticated malware has necessitated advanced detection techniques. This paper presents a novel deep learning approach for multi-class malware detection by leveraging a modified Graph Neural Network. Specifically, the deeperGcn,a cutting-edge deep learning model, is employed to enhance the feature extraction capabilities. Recognizing the imperative need for transparency in machine learning models, especially in cybersecurity, this approach integrates Explainable AI (XAI) principles using the GradCAM method, allowing for the interpretability of model decisions. A critical step in the methodology involved the merging of diverse datasets: both ASM (Assembly) and BYTE files. This comprehensive dataset amalgamation ensures the capture of intricate malware behaviors, thus bolstering the accuracy and reliability of the detection system. The approach strikes a balance between achieving high detection rates and accuracy upto 97% that stakeholders can intuitively understand and trust the model's predictions, paving the way for more robust and accountable malware detection systems in the future.