Secure AI SOC Workflows

Inspired by the CompTIA SecAI+ training certificate, this project bridges theoretical certification concepts with hands-on, defensible engineering. The primary objective is to demonstrate the secure deployment of a local Large Language Model tailored for modern SOC operations. At its core, the pipeline securely ingests analyst runbooks and utilises Retrieval-Augmented Generation (RAG) to automate live alert triage—testing how local LLMs can alleviate alert fatigue and streamline analyst workflows.

Beyond baseline automation, the project implements rigorous hardening and risk mitigation: evaluating LLM vulnerabilities against the OWASP Top 10 for LLMs and Generative AI applications, establishing strict access controls to secure sensitive knowledge bases, and showcasing practical AI governance including policy-as-code auditing, data minimisation, masking, and anonymisation.