Simulated: the people, messages, checks, and wearable are fake. Nothing leaves this page.
What you see
Trust meter
Pass 3 of 5 to proceed. No single check decides.
Five checks
What the enforcer sees
The app, the wearable vendor, and the identity provider each see only their part.
Liveness statement from the wearable:
Try this
- Pick Real friend. Click Run all five. One check fails and Sam still passes.
- Pick the cloned voice. Read why the code word passes and the score still stops it.
- Switch to Phone call. Pick Compromised real account. Press Verify.
Pain point
We cannot easily tell that the person who emails, texts, or calls is a real human and the one we expect.
Problem statement
A scammer can sound like someone you trust. One weak check can be faked.
What sparked it
Deepfake voice and helpdesk scams are widely reported.
Why this approach
Combine five weak checks into one score. Pass 3 of 5 and you proceed.
Hypothesis tested
We believed that five weak checks that agree are harder to fake than one. If it holds, 3 of 5 is a workable bar for a bank call.
How to try it
- Open the Prototype tab. Pick a surface and a scenario.
- Click Run check, or Run all five.
- Read the trust meter and the enforcer panel.
- Try the other scenarios.
What this does not do
- A pulse can be replayed. It proves a body, not an identity.
- No check helps if the attacker holds your devices.
- This is a simulation. No real wearable or call is used.
Notes
- Who would pay: banks, IT help desks, managed service providers.
- First customer: a managed service provider that resets passwords by phone.
- Closest alternative: one strong check, such as a passkey.
- Weak spot: three vendors must cooperate, and none wants to go first.
- Feedback: is 3 of 5 the right bar for a bank call?
Traces
| Step | Prompt | Model | Tool | Date |
|---|
Cost
| Step | Model | Input tokens | Output tokens | Cost (USD) |
|---|---|---|---|---|
| Total | ||||
Measured: the author runs used 103,218 tokens in total. Estimated: the split into input and output tokens (90 percent input, 10 percent output) and the dollar cost, at $2 input and $10 output per million tokens for Sonnet 5.5 (about $0.29). The run reported one total only. The lead session's planning, research, and publishing are not included. The prompts in the Traces section are listed in the order they were given.