Skip to content

ZoaGrad/sovereign-reliability-lab

Repository files navigation

Sovereign Reliability Lab

Cold Audit Node.js

Deterministic Fleet Synchronization • Forensic Verification • Reproducible Evidence

A research-grade experimental harness for proving deterministic workload coordination under stochastic network conditions. Produces auditable evidence artifacts instead of screenshots or dashboards.


What This Is

Most reliability demos visually “look stable” but cannot be independently verified. They smooth graphs, hide dropped events, and rely on trust in the UI.

Sovereign Reliability Lab is a deterministic synchronization simulator that generates tamper-evident forensic artifacts proving whether a coordination algorithm maintains temporal partitioning under noise.

It is designed to answer one question:

Can a distributed fleet preserve deterministic timing structure when subjected to realistic network jitter and packet loss?

The system produces machine-verifiable evidence that can be audited without trusting the UI.


What It Demonstrates

✔ Deterministic slot allocation using stable hashing ✔ Global barrier synchronization across a worker fleet ✔ Stochastic transport noise injection (Gaussian jitter + tail spikes) ✔ Structural integrity measurement (Spread-of-Arrivals Index) ✔ Persistence-gated resilience boundary detection ✔ Tamper-evident evidence export (canonical hashing) ✔ Independent CLI verification and report generation ✔ Full forensic replay from raw data


What It Does Not Claim

✘ Not a production scheduler ✘ Not a cryptographic security system ✘ Not a real network emulator ✘ Not optimized for high-frequency trading or sub-millisecond accuracy ✘ Not a consensus or distributed database

This is a measurement instrument, not a deployment artifact.


How It Works (High Level)

  1. A global barrier tick synchronizes a fleet of simulated workers.
  2. Each worker receives a deterministic time slot derived from a stable hash.
  3. Network noise is injected after scheduling (transport jitter + spikes).
  4. Arrival distributions are measured per tick.
  5. Structural quality is quantified using a normalized spread metric (SPI).
  6. A persistence gate promotes the maximum sustainable noise boundary (CSE).
  7. All raw samples, configuration, and metrics are exported as a signed artifact.
  8. A headless auditor recomputes every metric directly from the samples.

If the exported data is modified, verification fails.


Quickstart

Prerequisites

  • Node.js 18+

Install

npm install

Run the UI

npm run dev

Open the local URL shown in the terminal.


Run the Standardized Experiment ("Gold Run")

The system includes a deterministic 90-second experiment sequence.

  1. Click Start Gold Run (90s) in the UI.

  2. The system automatically:

    • Establishes a baseline.
    • Injects controlled network entropy.
    • Measures structural stability.
    • Generates an evidence artifact.
  3. Click Commit / Export to download the fingerprint JSON.

  4. Drag the file back into the UI to verify replay integrity.

This produces a portable proof artifact.


Verify Evidence via CLI (Headless)

All verification logic runs outside the browser.

Verify integrity

npm run audit:verify evidence/runs/fingerprint-XXXX.json

Exit codes:

  • 0 — Validated
  • 2 — Tampered
  • 3 — Evidence deficit

Generate formal audit report

npm run audit:report evidence/runs/fingerprint-XXXX.json

Outputs a Markdown Appendix A-1 audit report.


Repository Structure

cmd/                 # CLI auditor
utils/forensics.ts   # Canonical verification engine
components/          # UI visualization
services/            # Simulation logic
evidence/
  runs/              # Raw evidence artifacts
  reports/           # Generated audit reports
  protocol/          # Frozen forensic rules
test/                # Forensic unit tests

Reproducibility & Auditability

This project supports:

  • Deterministic experiment replay
  • Canonical serialization for tamper detection
  • Independent recomputation of metrics
  • CI-enforced artifact validation
  • Formal audit report generation

An auditor does not need to trust the UI.

They only need the JSON artifact and the verifier.


License

MIT

About

Audit-grade experimental harness for proving deterministic phase partitioning under stochastic network conditions. Includes reproducible evidence generation, cold verification tooling, and forensic replay.

Topics

Resources

License

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors