Institutional Repository of Vilnius University
Varžymosi principais grįstų atakų aptikimas naudojant paaiškinamo dirbtinio intelekto metodą kenkėjiškų programų obfuskacijos atveju /
Abstract
dc:descriptionThis work analyses adversarial attacks against malware detectors as well as strategies of detection and defense against such attacks. With the goal to improve already existing detection methods and adapt them for use with any feature vectors, this work explores combining MCA dimensionality reduction method and LIME – a method for explaining ML models’ decisions as these are implementations of the most perspective strategies of defense against adversarial attacks mentioned in scientific literature. A study is conducted to determine the accuracy of the suggested method by comparing it with the aforementioned strategies of defense against adversarial attacks. MalGAN framework is used for this study as the generator of such attacks since the binary vectors used by it should cause the most difficulty for existing strategies of defense. Experimental results show that the suggested method is able to both defend against and detect adversarial attacks more effectively and consistently than currently used methods.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kasperavičius, Liudas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD239030996&prefLang=en_US