GNN-Based Detection of XSS Vulnerabilities to Strengthen Security for Financial Web Transaction: A Web Browser Extension Approach

DeepLearningIndaba 2025 Workshop AIBF Submission8 Authors

Keywords: Ai-Driven for Security, Cross-Site Scripting (XSS), Online Business security, Graph Neural Networks (GNN), Web Malicious Code Detection

TL;DR: AI-Powered Detection of Malicious Script to enhance cybersecurity for Business and Financial Web Applications

Abstract: The rise of digital financial platforms in Africa has improved access to services but also increased the risk of cyberattacks like Cross-Site Scripting (XSS). XSS attacks inject dangerous JavaScript code into websites, which can steal user data or cause other harm. To help protect users, we developed a Firefox extension that detects malicious scripts in real time. This extension uses a Graph Neural Network (GNN), a type of AI model we have already trained on Control Flow Graphs (CFGs), to find hidden and complex malicious code. The extension shows a clear alert when it detects suspicious code and highlights the dangerous part. It also includes a chatbot assistant based on ChatGPT that explains the code’s behavior in simple words. We tested the extension on real African financial websites and with sample data, and it showed good results. This tool combines strong AI detection with easy explanations to improve online safety and user awareness.

Submission Number: 8