O.M.E.G.A
A unified Active Directory workspace for guided administration, operator-based permissions, and traceable enterprise identity operations.
Explore O.M.E.G.ASecurity research
RAVEN
A multi-agent security research engine that connects parallel investigations, target-specific testing, and benchmark evidence.
Engineering illustrationHYPOTHESIS · TEST · EVIDENCE
Project status
CVE benchmarking
Technology stack
Engineering focus
From a security hypothesis to testable evidence.
RAVEN connects exploratory reasoning with concrete testing. AI agents investigate a target, form hypotheses about possible vulnerabilities, and write tests tailored to those hypotheses—bringing the investigation and its evidence into one research workflow.
01 / Context
A suspected weakness needs a testable explanation.
Security research becomes useful when a suspicion can be turned into a concrete test. RAVEN explores that transition: agents examine the target, develop competing hypotheses, and investigate the behavior that would support or challenge those hypotheses.
02 / The system
Multiple investigations. Tests shaped by the target.
RAVEN coordinates multiple AI agents across codebases, repositories, and web applications. Agents form vulnerability hypotheses, write tailored attack tests, and run those tests as part of the investigation. The project’s current CVE benchmarking phase is intended to measure detection performance before its planned open-source publication.
03 / Capabilities
Research that connects reasoning and execution.
Parallel research paths
Multiple AI agents investigate different hypotheses about a codebase or application.
Tests shaped by the target
The platform writes tests for the specific weaknesses under investigation, connecting reasoning with executable validation.
Benchmark-driven development
CVE benchmarking provides a framework for evaluating detection performance. Publication is planned after the project’s benchmarking goals are met.
04 / System map
A research loop built around evidence.
This conceptual sequence explains how the project’s main capabilities connect.
Examine the target
Agents investigate the supplied codebase, repository, or web application.
Form hypotheses
Different investigation paths propose possible weaknesses.
Build targeted tests
Tests are written and run for the hypotheses being explored.
Assess the evidence
Results feed the investigation and the ongoing benchmark evaluation.
05 / Scope & status
Current project scope.
The platform is built and undergoing CVE benchmarking. Open-source publication is planned after its benchmarking goals are met.
For questions about this project or its relevance to your organization, contact admin@psiphas.org.
Continue exploring
Related systems.
AD API Center
A shared authentication foundation for internal applications, connecting Active Directory identity with origin-bound profiles and application-specific access rules.
Explore AD API CenterBuild on a solid foundation
Get in touch