Institutional Repository of Vilnius University
Įterptinių komandų atakų generavimas naudojant didžiuosius kalbos modelius interneto technologijų saugumo kontekste /
Abstract
dc:descriptionCross-Site Scripting (XSS) vulnerabilities are becoming a more common method of attacking web applications due to their relative simplicity and considerably severe effects when successfully executed. To properly defend against the immense number of XSS attack variations, ethical security specialists may benefit from dynamically generated XSS strings for use in security testing on applications they are protecting. This research explores the capabilities of Large Language Models to generate XSS attack strings, analyzes different methods and techniques of exploiting XSS vulnerabilities. Two LLM models - Code Llama and Llama 3 are fine-tuned using a dataset, containing different examples of JavaScript code, used for XSS attacks. After proving the existence of a vulnerability in an ethical hacking training site, same vulnerability is attacked by LLM-generated strings using Burp Suite penetration testing tool. Results show that while the majority of Llama 3 generated XSS strings does not exploit the vulnerability, Code Llama shows much more promising results, successfully exploiting aforementioned vulnerability.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Anikanovas, Martynas,
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:ELABAETD210644775&prefLang=en_US