Cross-site scripting adversarial attacks based on deep reinforcement learning: Evaluation and extension study

TL;DR

This paper evaluates and extends a deep reinforcement learning-based adversarial attack on XSS detection models, demonstrating a high escape rate and proposing an XSS Oracle to improve robustness assessment.

Contribution

It replicates and critiques a state-of-the-art XSS adversarial attack, extends its evaluation, and introduces an XSS Oracle for more reliable security testing.

Findings

Achieves over 96% escape rate against detection models

Highlights threats to validity in existing adversarial attack methods

Proposes an XSS Oracle to improve evaluation robustness

Abstract

Cross-site scripting (XSS) poses a significant threat to web application security. While Deep Learning (DL) has shown remarkable success in detecting XSS attacks, it remains vulnerable to adversarial attacks due to the discontinuous nature of the mapping between the input (i.e., the attack) and the output (i.e., the prediction of the model whether an input is classified as XSS or benign). These adversarial attacks employ mutation-based strategies for different components of XSS attack vectors, allowing adversarial agents to iteratively select mutations to evade detection. Our work replicates a state-of-the-art XSS adversarial attack, highlighting threats to validity in the reference work and extending it towards a more effective evaluation strategy. Moreover, we introduce an XSS Oracle to mitigate these threats. The experimental results show that our approach achieves an escape rate above 96% when the threats to validity of the replicated technique are addressed.