Process Improvement and Optimization of Neuro-Symbolic Expertise for Enhanced RMF/ATO (PIONEER)

The Authority to Operate (ATO) process under the DoD Risk Management Framework (RMF) is critical for cybersecurity but is slow, resource-intensive, and requires deep expertise. Security teams must manually align complex systems with evolving regulations, leading to delays and inconsistencies that hinder mission readiness. To address these needs, Charles River Analytics proposes to design and demonstrate the feasibility of Process Improvement and Optimization of Neuro-Symbolic Expertise for Enhanced RMF/ATO (PIONEER), a neuro-symbolic AI-driven system that leverages subject matter expert (SME) mental models to streamline and automate the creation of RMF-compliant ATO packages. PIONEER reduces time required to generate products and ensures consistency through a combination of three innovative components. First, Symbolic Mental model Automation and Reasoning Technology for Enhanced RMF/ATO (SMARTER) processes encodes SME reasoning into structured, machine-readable representations, allowing AI to assist in RMF/ATO compliance decision-making and documentation. Second, large language model (LLM) GRammar-Assisted Fine-Tuning (GRAFT) techniques enhance LLMs through fine-tuning, adaptive knowledge integration, and expert-guided adaptation, ensuring that AI responses align with domain-specific compliance requirements and evolving policies. Third, LLM External Tools Integration (LETI) connects LLMs with external cybersecurity tools, such as vulnerability scanners and configuration analyzers, enabling AI-driven risk assessments based on real-time system data rather than static documentation. By combining neuro-symbolic AI with adaptive learning and tool-based automation, PIONEER accelerates ATO approvals, reduces manual workload, and improves compliance accuracy, strengthening the Navy’s cybersecurity posture while ensuring that mission-critical systems achieve compliance efficiently.