Author(s)
Prof. Sayali Gaikwad, Siddhi Ghogare, Tejaswini More, Nikita Chitalkar, Trupti Ganeshkar
- ISSN (P): 3139-8464
- Manuscript ID: 141104
- Volume: 2
- Issue: 10
- Pages: 574–580
Subject Area: Computer Science and IT
Abstract
Web applications are increasingly exposed to security vulnerabilities such as SQL injection, cross-site scripting, command injection, and insecure access-control weaknesses. Traditional Static Application Security Testing (SAST) and Dynamic Application Security Testing (DAST) techniques provide important capabilities for identifying such vulnerabilities, but the subsequent analysis and remediation process can still require substantial developer effort. Recent research has introduced artificial intelligence, machine learning, deep learning, and large language models (LLMs) for vulnerability detection, prioritization, autonomous security assessment, and automated code remediation. However, existing studies often address these capabilities as separate tasks, while practical integration of detection, code-context extraction, patch generation, patch validation, and repository-level remediation remains a significant challenge. This review examines the evolution of AI-assisted web application vulnerability detection and automated remediation and analyses representative approaches in the existing literature. Based on this analysis, the paper discusses AutoPatch AI as an integrated workflow that connects SAST/AST and DAST-based vulnerability detection with code-context extraction, LLM-based patch generation, isolated Docker-based validation, post-patch re-scanning, code-diff generation, and developer-approved GitHub or GitLab pull requests. The review further identifies key challenges related to patch reliability, validation, generalization, benchmarking, and human oversight. The proposed workflow provides a practical direction for connecting vulnerability discovery with secure and verifiable code remediation within a DevSecOps-oriented development process.