Skip to content

Research

Research SYMMACHY conducts in public.

Default is publish. Proprietary requires written justification, signed by a founder. The full R&D framework lives in the corp-handbook-docs repository. The streams below describe the research direction; the publication record will appear on this page as it accumulates.

Stream A

AI Security & Frontier Risk Evals

Adversarial machine learning at runtime — prompt injection, jailbreak, indirect injection, multi-modal evasion, agent hijacking. Continuous frontier-risk evaluations against deployed models.

Planned outputs: One to two peer-reviewed papers per year (target venues: IEEE S&P, USENIX Security, NDSS, ACM CCS).

Stream B

Cyber-Physical Fusion

Sensor-fusion ontologies for cyber and physical telemetry under one second of end-to-end latency. Graph-temporal reasoning over heterogeneous evidence.

Planned outputs: Reference architecture document and a phish-to-substation-breaker benchmark.

Stream C

Robotic Perception & Edge AI

Foundation models for robots running on Jetson Thor at fifty hertz. Customer digital twins via NVIDIA Cosmos and Isaac Sim.

Planned outputs: Open-source perception library; hardened Vision-Language-Action defence pipeline.

Stream D

Evidence, Cryptography, Identity

Sigstore extended for physical sensor telemetry. SPIFFE identities that attest hardware integrity. Post-quantum-ready signing.

Planned outputs: SYMMACHY-extended Sigstore reference implementation; a public PQC migration plan.

Published works

This section lists SYMMACHY publications in chronological order, each linked to its venue (arXiv, conference proceedings, GitHub release) and to its evidence-chain anchor in Rekor.