Secure Enclave Design
Overview
Section titled “Overview”The Secure Enclave Pattern enables intra-tenant connected context. The intent is to securely enable the discovery of cross-data relationships, even across security boundaries.
Imagine a scenario where Alice is an engineer working on the public OpenShift Console repository. She is investigating a memory leak.
There is a known, but embargoed security vulnerability (CVE) that identifies the root cause of this leak in a specific upstream library, npm:axios.
Alice does not have read access to the embargoed CVEs. However, the relationships in Kartograph allow her AI agent to discover the dependency without revealing the sensitive details of the vulnerability itself.
The AI agent can see that openshift/console is affected by a cve that has a root cause in npm:axios, but sensitive information like the CVE number
and vulnerability details are redacted. The agent only sees the non-identifying secure ID of the CVE (4418c846ef82aeef, a hash of CVE-2025-01).
graph LR
A([openshift/console]) -- Affected By --> B["🔒 cve:4418c846ef82aeef"]
B -- Root Cause --> C([npm:axios])
style A fill:#e6f3ff,stroke:#333,stroke-width:2px
style B fill:#ffe6e6,stroke:#ff0000,stroke-width:2px,stroke-dasharray: 5 5
style C fill:#e6f3ff,stroke:#333,stroke-width:2px
How It Works
Section titled “How It Works”Kartograph provides secure connected context via:
- Database-level tenant isolation prevents malicious queries from leaking data
- Native classification controls at the data source level allow access-based content filtering
- Tenant-scoped secure IDs for nodes and relationships prevent correlation between tenants in the event of a zero-day that enables queries to query across databases
- This can also enable secure deterministic inter-graph linking of nodes.
- ReBAC-based post-query redaction of data stored in the secure enclave ensures clients only receive data they are cleared to see