People type their own name, birthday and ZIP code from motor memory. Someone using a stolen identity has to read, recall or compute each answer, and that leaves gaps in the keystroke rhythm. Fill in both forms below and compare your two rhythms.
● Only key timings are measured, never which keys. Nothing leaves this page.
1 · Your identity
familiar
Use your own details, or any identity you have typed hundreds of times. Values are never stored.
2 · An assigned identity
unfamiliar
First nameDelphineLast nameOkonkwo-HartDate of birth11/23/1979ZIP code58203
Type it; don't paste. The follow-up is an "unexpected question": the owner knows it, an impostor has to work it out.
Type at least a few characters in every field of both forms.
Readout
Example session · yours replaces it
Your identityAssigned identityEach tick is one key press; gaps are hesitations. Field boundaries are marked below each line.
From the replication
The step-up budget
A score does not convict anyone. A platform sends the riskiest share of new applicants to a slower check (document scan, selfie, call-back) and lets everyone else through. These curves come from re-running Monaro et al.'s published classifier on their 151 online participants. Pick a fraud rate and a budget and read off what the screen buys.
Model
of fraud caught
of step-ups are fraud
of genuine users skip friction
Fraud caught (vertical) against share of all applicants sent to step-up (horizontal), at the chosen fraud rate. Ties in the score are broken at random. Trained on 40 lab participants, tested on 151 recruited online.