EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Creating valid adversarial examples of malware

Matouš Kozák, Martin Jureček, Mark Stamp, Fabio Di Troia

Journal of Computer Virology and Hacking Techniques · 18.03.2024 · 20 (4) · 607-621

DOI: 10.1007/s11416-024-00516-2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Computer Virology and Hacking Techniques, online erschienen am 18.03.2024. Verfasst von Matouš Kozák und 3 weiteren Autor:innen. Die Arbeit wurde bisher 34-mal zitiert (Crossref) und nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministerstvo Školství, Mládeže a Tělovýchovy, Research Center for Informatics, Czech Technical University in Prague.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Because of its world-class results, machine learning (ML) is becoming increasingly popular as a go-to solution for many tasks. As a result, antivirus developers are incorporating ML models into their toolchains. While these models improve malware detection capabilities, they also carry the disadvantage of being susceptible to adversarial attacks. Although this vulnerability has been demonstrated for many models in white-box settings, a black-box scenario is more applicable in practice for the domain of malware detection. We present a method of creating adversarial malware examples using reinforcement learning algorithms. The reinforcement learning agents utilize a set of functionality-preserving modifications, thus creating valid adversarial examples. Using the proximal policy optimization (PPO) algorithm, we achieved an evasion rate of 53.84% against the gradient-boosted decision tree (GBDT) detector. The PPO agent previously trained against the GBDT classifier scored an evasion rate of 11.41% against the neural network-based classifier MalConv and an average evasion rate of 2.31% against top antivirus programs. Furthermore, we discovered that random application of our functionality-preserving portable executable modifications successfully evades leading antivirus engines, with an average evasion rate of 11.65%. These findings indicate that ML-based models used in malware detection systems are sensitive to adversarial attacks and that better safeguards need to be taken to protect these systems.

Förderung

  • Ministerstvo Školství, Mládeže a Tělovýchovy
  • Research Center for Informatics, Czech Technical University in Prague

Zitieren

APA

Kozák, M., Jureček, M., Stamp, M., & Troia, F. D. (2024). Creating valid adversarial examples of malware. Journal of Computer Virology and Hacking Techniques, 20(4), 607-621. https://doi.org/10.1007/s11416-024-00516-2

Vancouver

Kozák M, Jureček M, Stamp M, Troia FD. Creating valid adversarial examples of malware. Journal of Computer Virology and Hacking Techniques. 2024;20(4):607-621. doi:10.1007/s11416-024-00516-2

BibTeX
@article{Kozk2024,
  author = {Kozák, Matouš and Jureček, Martin and Stamp, Mark and Troia, Fabio Di},
  title = {Creating valid adversarial examples of malware},
  journal = {Journal of Computer Virology and Hacking Techniques},
  year = {2024},
  volume = {20},
  number = {4},
  pages = {607--621},
  doi = {10.1007/s11416-024-00516-2}
}
Kostenlos starten →