TrueLoop x PQC-QKD: reproduction code

Reproduce the PQC-QKD physical-layer stabilization results using the live TrueLoop runtime endpoint. These scripts drive the real runtime (one measurement per round, model-free) against a simulated QKD/QRNG plant. Swap the simulated plant for your hardware to validate on a real entropy source or QKD link.

What this is (and what it is not)

• It is the exact experiment code that produced our simulation results, calling the public runtime endpoint. You send a measurement each round and receive the next control configuration.
• It is not the runtime's internal method. That runs server-side and is not in this code. These scripts only use the runtime's measured input/output behaviour.

Claim tier (please preserve this when sharing results)

All numbers these scripts produce are SIMULATED PLANT + LIVE RUNTIME. They demonstrate the control law's behaviour, not validation on real QKD hardware. Hardware validation is the open question this code is meant to help answer.

Setup

1. Get a free evaluation key at https://compute.neophotonics.ca/
2. Open trueloop_client.py and replace YOUR_EVAL_KEY with your key.
3. Check connectivity:

   python trueloop_client.py
   

   You should see OK - endpoint reachable and key valid.

Requires only Python 3 standard library (no dependencies).

Run the experiments

python experiment_qkd_keyrate.py        # headline: secure key rate + link uptime under drift
python experiment_qkd_suite.py A        # QRNG entropy bias stabilization
python experiment_qkd_suite.py B        # QKD modulator/interferometer bias lock
python experiment_qkd_suite.py C        # polarization / state-alignment hold
python experiment_attack_safety.py      # runtime does not mask an attack's QBER signature

Validate on YOUR hardware

Each experiment isolates the simulated plant in one place, marked hardware swap point / measure_plant() / measure(). To validate on real hardware:

1. Replace that function so it returns the per-channel error signal read from your real device under the current control configuration phi.
2. Apply the configuration the runtime returns (r["phi"]) to your real control variables (modulator bias, phase shifter, polarization controller, etc.).
3. Keep the loop structure identical: measure once, send, apply the returned config, repeat.

The runtime never needs a model of your device; it discovers the control directions from your measurements.

Known envelope (where this is expected to work)

• Works: slow-to-moderate drift on a single link where each measurement is costly.
• Does not scale with channel count (the per-link QKD objective is forgiving).
• Bandwidth / capture-range limited: degrades under fast structured drift or large sudden excursions; pair with coarse re-acquisition for big jumps.

Contact

https://trueloopcompute.com · https://compute.neophotonics.ca/