Born Between 2Generals
← Showcase homePage status: Research · synthetic demo only

Verify · Research

Checking that a person is present — without overclaiming

A research design for voluntary, consented, one-to-one check-in. It combines cardiac and optical pulse measurements with document, credential and session checks — and keeps every result honest about what it cannot know.

Start the synthetic walkthrough See all 21 checks

Research only. Nothing here identifies anyone. The demonstration uses synthetic signals, its matcher returns not_evaluated, and every receipt states valid_for_authentication: false. No accuracy figures exist and none are claimed. This must never become a condition for voting.

The core idea

Four decisions, kept separate

None of these establishes the others. A real, living person can still present someone else's identity, and a correct identity match does not establish eligibility for anything.

Four separate decisions Presence: is there evidence of a live person at this session? Identity: does that person match the claimed, properly enrolled identity? Session integrity: did fresh evidence come from the expected devices without substitution? Eligibility or authorisation: does the responsible authority permit this action? Each is decided separately; the fourth belongs only to the responsible authority. 1 · Presence Is there evidence of a live personat this capture session? 2 · Identity Does that person match the claimed,properly enrolled identity? 3 · Session integrity Did fresh evidence come from the expecteddevices, without substitution? 4 · Eligibility / authorisation Decided only by the responsible authorityunder its own rules — never by this system.

Walkthrough video

See the synthetic session

A silent recording of the browser walkthrough: consent, synthetic enrolment, five capture windows (one fails its quality check and is retried), an honest “inconclusive” comparison, and a test receipt that cannot authorise anything.

  1. Consent, with the non-biometric alternative offered first.
  2. Five capture windows: quiet baseline, two neutral prompts, a quiet pause and a final confirmation.
  3. Comparison over all usable windows by a fixed rule — never the best single match.
  4. A minimal test receipt, marked simulated and not valid for authentication.

Silent screen recording with captions.

Read the transcript
  1. 00:00 This is the synthetic verification walkthrough. It uses no camera, microphone or sensor, and nothing is saved.
  2. 00:03 Consent comes first, and the non-biometric alternative is offered before anything else.
  3. 00:07 Synthetic enrolment: a service-specific pseudonym, a synthetic identity review and a test-only credential.
  4. 00:11 Five capture windows follow. Each is checked for quality before anything else.
  5. 00:15 Window two asks the participant to read a fresh number. Typing it is an accessible alternative.
  6. 00:19 That attempt failed its quality check because of motion. A capture problem is not a judgement about the person.
  7. 00:22 A bounded retry succeeds. The flow channel stays “unavailable”: there is no qualified hardware.
  8. 00:27 All usable windows are combined by a fixed rule. The matcher returns “not evaluated”, so the result is honestly inconclusive.
  9. 00:32 The person approves exactly one synthetic action. It cannot register, verify or authorise anything.
  10. 00:36 The test receipt holds no biometrics and no ballot choice, and says valid for authentication: false.
  11. 00:41 Redeeming it twice is rejected as a replay.
  12. 00:44 A production verifier rejects it outright: test issuer, simulated evidence.

Commitments

What the research will not do

  • No public-camera identification, covert location searches or global tracking index.
  • No lie detection: a physiological reaction is not evidence of dishonesty.
  • No automatic fraud labels, roll removals or denials from a failed sensor or an unanswered message.
  • No biometric data in ballots, receipts or audit exports, and no link between identity and ballot choices.
  • No treating five windows of one session as five independent proofs.

Before any real use

Gates that must be passed first

  • Qualified hardware with raw-data access and safety documentation.
  • A reviewed adult volunteer study with repeat visits, designed with a statistician.
  • Predefined error, capture-failure and attack-acceptance measures, reported with uncertainty.
  • For any election pilot: approval by the responsible authority, jurisdiction-specific legal review and an accessible non-biometric route.

Next step: read the 21 checks or see the election evidence programme.

Behind this page

A fuller research prototype runs privately

Private · local only The BB2G Verification Lab (version 0.1.0) is a synthetic-only working prototype on the BB2G Probox. It includes the enrollment simulator, five-window capture, the 21-check registry, test-only signed receipts, a review queue and an audit room. It is not published and has no public address; this page is its public explanation.

Its own tests run on synthetic data only. Passing them does not validate any biometric method, sensor or election use.