MiBeX: Malware-inserted Benign files for Explainable Machine Learning

 

Abstract

This work explores the use of Metasploit to produce a scalable malware dataset with intelligible malicious features for machine learning (ML) classification. The resultant dataset is then used with a Malware as Image classifier to prove its validity for use in training ML algorithms. A dataset containing 2206 files is generated, and a ML classifier achieves 99.72% test accuracy on the dataset.

​

Poster