An AI-Enhanced Framework for Scalable Security Architecture Analysis
S. C. Das Gupta, Anupam Mehta · International Journal For Multidisciplinary Research · 2025
Traditional threat modeling approaches often fail to scale in modern DevSecOps environments due to their reliance on static analysis, manual processes, and limited developer integration. We present an AI-enhanced framework that reimagines threat modeling as a continuous, real-time process embedded within CI/CD pipelines. Leveraging large language models, the framework extracts architectural insights from source artifacts—such as Terraform, GitHub, and OpenAPI specs—to detect risks, infer trust boundaries, and prioritize threats based on exploitability and business context. Rather than performing one-time reviews, it synchronizes with ongoing development, automatically tracking architectural drift and surfacing contextualized security insights. Evaluation across five diverse software architectures—including microservices, serverless functions, and ML APIs—demonstrated measurable improvements, including an 83% reduction in modeling time and a 26% increase in critical threat detection. These results suggest that AI-assisted threat modeling can provide scalable, developer-aligned security design without compromising speed or precision.