← Back to projects
Cybersecurity — AI automation
VulnAI
PythonOllama / MistralNVD APIMITRE ATT&CK
Context
Manually tracking CVEs across an application estate is time-consuming and poorly prioritized: it's hard to quickly tell which flaws pose a real risk to the organization.
What was done
Built a Python pipeline that queries the NVD API, cross-references detected CVEs with the corresponding MITRE ATT&CK techniques, then uses a local LLM (Ollama / Mistral) to generate a prioritized risk summary understandable without deep security expertise. The pipeline was validated end-to-end, up to confirming an RCE (uid=0).
Result
Confirmed remote code execution in a test environment (uid=0), demonstrating the relevance of the pipeline's automated risk prioritization.