Hacking websites with Reinforcement Learning: an XSS story
Lightning Talk
Reinforcement Learning; Cross Site Scripting (XSS)
Broken authorisation leads to some of the most common vulnerabilities in web APIs. Using automated testing to find these vulnerabilities is difficult due to dependencies on access control policy, and the specific vulnerability. In this talk, we discuss how a reinforcement learning implementation can be developed to maximise the potential to find broken authorisation vulnerabilities in web APIs. We will focus on the specific problems associated with forming it into an RL game, and generalisation to multiple APIs. The selected implementation is then tested against to confirm its ability to find vulnerabilities.