Let's visualize malwares ! Malware visualization and classification

Details

Supervisors

Legay, Axel

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en informatique, à finalité spécialisée

Abstract

Due to the growing threat from malicious softwares, security of sensitive networks is becoming increasingly crucial. The ever-growing number of suspicious samples is a time-consuming process for malware analysts and the need is real for a tool able to quickly analyze a large number of samples. The combination of new visualization methods from dense pixel displays based on raw sample data, combined with promising technologies for the image classification such as convolutional neural networks, is able to meet this need. In this master thesis, we provide two new visualization techniques and a classifier able to categorize new samples into malware families at the entry of a high-speed network with an accuracy up to 97.82%. In case of doubt, samples can still be analyzed in more depth by malware experts. Additionally, we explore different datasets, the impact of their structure on classification accuracy and we analyze the insertion of cleanwares in the dataset. The whole implementation of our work is freely accessible on GitHub: https://github.com/Tioneb88/TFE_Malware_Visualization_and_Classification