security-hive / README.md
Nicholastempleman's picture
Agent: security-hive README — the nightly attack loop, published
257dd9e verified
|
Raw History Blame Contribute Delete
1.74 kB
---
license: cc-by-4.0
language: [en]
pretty_name: Security hive — nightly attack loop, signed
tags: [gspc, council-of-ai, security, serving-fidelity, owasp-asi, signed-evidence]
size_categories: [n<1K]
---
<p align="center">
<img src="https://huggingface.co/datasets/csoai/brand-assets/resolve/main/csoai-banner-1280x256.png" alt="Council of AI (CSOAI)" width="100%">
</p>
# Security hive — the nightly attack loop, published
Every night at 04:23 UTC, an attack harness (`hive-attack-loop.mjs`, sha256-pinned in each record)
fires **33 attacks** at the loopback and LAN interfaces of the running MEOK product — then records,
signs (Ed25519), and timestamps the result. **These are those signed records.**
| Field | Meaning |
|---|---|
| `axis` | The GSPC axis this nightly run measures (`serving-fidelity`) |
| `method.attacks` | 33 attacks per run, expectations measured against the running product beforehand |
| `result` | Held / failed per attack — latest runs: **31/31 held** |
| `honesty` / `limitations` | What this run does and does not claim |
| `signature` | Ed25519 under the published DID key set — verify at [gspc-verify](https://councilof.ai/gspc-verify) |
Also included: the **OWASP ASI → GSPC axis map** (the OWASP Agentic AI threats, mapped onto GSPC
measurement slots; measured 10/10 at 0% false positives, 2026-10-27).
**Why this exists:** a security posture that only lives in a private repo is a claim. Signed nightly
records in public are a measurement. Serving-fidelity — the axis this hive measures — is currently a
declared-UNMEASURED board slot elsewhere; this hive is the evidence source for flipping it.
Measurement, not certification. Free verification: https://councilof.ai/gspc-verify