Document Type



Master of Science (MS)


Computer Engineering

First Advisor's Name

Dr. Selcuk Uluagac

First Advisor's Committee Title

Committee Chair

Second Advisor's Name

Dr. Kemal Akkaya

Second Advisor's Committee Title

Committee member

Third Advisor's Name

Dr. Alexander Perez-Pons

Third Advisor's Committee Title

Committee member


P.E. malware, cryptojacking, deep-learning, adversarial machine learning, web assembly

Date of Defense



This graduate thesis introduces novel, deep-learning based frameworks that are resilient to adversarial P.E. and cryptojacking malware. We propose a method that uses a convolutional neural network (CNN) to classify image representations of malware, that provides robustness against numerous adversarial attacks. Our evaluation concludes that the image-based malware classifier is significantly more robust to adversarial attacks than a state-of-the-art ML-based malware classifier, and remarkably drops the evasion rate of adversarial samples to 0% in certain attacks. Further, we develop MINOS, a novel, lightweight cryptojacking detection system that accurately detects the presence of unwarranted mining activity in real-time. MINOS can detect mining activity with a low TNR and FPR, in an average of 25.9 milliseconds while using a maximum of 4% of CPU and 6.5% of RAM. Therefore, it can be concluded that the frameworks presented in this thesis attain high accuracy, are computationally inexpensive, and are resistant to adversarial perturbations.







Rights Statement

Rights Statement

In Copyright. URI:
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).