This research will explore the capabilities of Large Language Models (LLMs) in generating complex, full-scale web applications using widely adopted AI web development frameworks (eg.: OpenDevin, MetaGPT, etc...). The study will focus on assessing the quality, usability, and security of AI-assisted web applications.
Objectives:
- Framework-Based Web Development: Use generative AI assisted frameworks to create large, functional websites. The aim is to examine how well AI can manage framework-based development, considering real-world requirements like database integration, authentication, and user management etc...
- Security Vulnerability Analysis: Conduct a thorough security analysis using automated tools (eg, Burp Suite, OWASP ZAP) and manual penetration testing techniques to identify vulnerabilities such as SQL Injection, XSS, CSRF, and other vulnerabilities. The evaluation will also include static code analysis to identify common issues in AI-assisted code.
- Comparative Study: Compare the different frameworks in terms of which approach creates the most secure websites.
Potential Research Questions:
- How effectively can AI generate complex, secure, and usable web applications using modern frameworks?
- What are the common security vulnerabilities in AI-generated large-scale websites?
- What methodologies can be developed to benchmark the quality and security of AI-assisted web applications?
- How do the different frameworks/AI-models compare in terms of secure code generation?