Locally Deployed NLP Models for Secure and Customizable AI-Enabled Common Criteria Evaluations (A31c)
This talk proposes a novel AI-enabled expert system for analyzing Common Criteria Evaluation documentation. Leveraging open-source NLP models and local execution, the system ensures confidentiality while utilizing models fine-tuned for Common Criteria standards. A lightweight web server facilitates programmatic access, enabling seamless integration with existing workflows. This approach enhances customization, confidentiality, and evaluation efficiency, offering a robust solution for rigorous Common Criteria evaluations.