PULSE Lab Selected for ReliaQuest Innovation Challenge Fund: Federated Agentic Access for Cybersecurity Knowledge Graphs

We are excited to share that the PULSE Lab was selected for the ReliaQuest Innovation Challenge Fund at Florida State University!

Our project

Federated Agentic Access for Cybersecurity Knowledge Graph Integration

• Challenge area: Challenge Three – Knowledge Graph Integration
• PI: Dr. Olmo Zavala-Romero (Scientific Computing)
• Project personnel: Brendon DGR, graduate research assistant / technical lead
• Amount requested: $29,298
• Project period: June 8, 2026 – October 23, 2026
• Pitch event: ReliaQuest Innovation Challenge, October 19, 2026 (FSU Discovery Days)

This seed award is one of up to six ReliaQuest Innovation Challenge Fund projects supporting proof-of-concept work in AI and cybersecurity.

What we are building

Cybersecurity teams increasingly rely on knowledge graphs for adversary techniques, attack patterns, mitigations, and threat intelligence — but those sources often live in different systems with different schemas and query languages.

Our project will implement a federated agentic graph-database access prototype grounded in two public cybersecurity knowledge graphs:

1. MITRE ATT&CK Enterprise STIX in Neo4j / Cypher
2. MITRE CAPEC ATT&CK-related attack patterns in a Gremlin Server environment approximating Amazon Neptune-style traversal

An orchestrator agent will route parts of an analyst’s natural-language question to source-specific query agents. Those agents generate validated graph queries, convert results into grounded text, and pass evidence to a synthesis agent — with no vector index or embedding store built from the graph databases. The focus is native, read-only access through Cypher and Gremlin, with provenance retained behind every claim.

Example: for a question like “Which CAPEC patterns related to credential theft map to ATT&CK techniques, and what mitigations appear for those techniques?”, the planner emits CAPEC and ATT&CK tasks, specialists run bounded Cypher/Gremlin queries, and synthesis joins on ATT&CK IDs to return a provenance-linked answer, table, and compact CAPEC–ATT&CK–mitigation view.

Deliverables by October 2026

By the ReliaQuest pitch, the team aims to deliver a working demonstration that:

• Queries ATT&CK and CAPEC as distinct graph-database sources
• Correlates attack patterns with ATT&CK techniques and mitigations
• Evaluates routing, query translation, answer grounding, and provenance correctness
• Documents guidelines for migrating the same agent framework to other query languages

A later stage could extend the same agent layer to managed graph backends such as Amazon Neptune and Google Cloud Spanner/Spanner Graph (openCypher, Gremlin, GQL).

About the ReliaQuest–FSU partnership

This work is part of ReliaQuest’s broader AI and cybersecurity partnership with FSU, backed by a $1.5 million philanthropic investment that established the annual ReliaQuest Innovation Challenge. Seed awardees develop prototypes over the summer and early fall, then pitch for follow-on prizes totaling up to $200,000.

Read more: ReliaQuest and Florida State University Launch AI and Cybersecurity Partnership.

Program details: FSU Office of Research – ReliaQuest Innovation Challenge Fund.