Malware classification using machine learning methods

OData support
Supervisor:
Dr. Hullám Gábor István
Department of Measurement and Information Systems

The detection of malware, also known as malicious software, is one of the most important aspects of the software products designed for protecting information systems. The endless variety of possible malicious activities, as well as their continuous adaptation to circumvent the defensive measures puts severe stress on the whole industry dedicated to malware defence, pushing it to develop always more advanced methods with the goal of creating a more efficient protection.

The fast-paced obsolescence of the known methods demands the search for ever newer solutions. Malware programs have evolved significantly since their early forms – worms, trojans, which spread with simple self-replication, today’s malware is more resembling to the current distributed cloud systems, which become ever more resistant to traditional and known detection techniques.

In this work, I’ve implemented an idea from a newer segment of the artificial intelligence scene - artificial intuition. This new technique has many advantageous properties in comparison to more established and used methods. The method’s description is novel in the field of malware detection, providing a possible breakthrough in the field, especially with comparing its possible capabilities to previous machine learning methods used.

For comparison, I’ve used the Microsoft Malware Classification (BIG 2015) challenge from Kaggle. which contains data for 9, behaviorally as well as structurally different malware classes, with classification as its stated goal.

Downloads

Please sign in to download the files of this thesis.