Multi-modal Machine Learning Model for Interpretable Malware Classification

As mobile devices have become universal and are now integral to every facet of our everyday lives, the alarming rise in mobile malware poses a significant threat to the security of sensitive and private information stored or transferred to/from our mobile devices (e.g., smartphones, and tablets). This paper introduces an innovative method for mobile malware detection using a multimodal deep learning approach on two different modalities of datasets: grayscale images of android malware and tabular data. We leverage Explainable AI (XAI) to enhance the interpretation of classification results for both unimodal and multimodal approaches. Furthermore, we create an explainable malware classifier using Knowledge Graph to compare its performance with multimodal learning. The classifiers provide improved explainability with minimal to no compromise in accuracy when classifying malware samples.