QSIdentify reports confidence as independent dimensions: transport, protocol, firmware, fingerprint, statistical, review and catalog confidence. These dimensions are never averaged or collapsed into a single confidence value.
Project thresholds are policy gates, not scientific proof:
- one independent radio supports stability analysis;
- two support comparison;
- three permit candidate correlation;
- five independently inspected radios permit human review;
- ten independently verified radios permit a catalog proposal.
Repeated captures from one physical radio are not independent devices. Statistical correlation can identify a useful discriminator candidate, but it cannot establish a hardware identity by itself. Physical inspection, electronic observations and reviewer conclusions remain visibly separate. Unknown or contradictory responses are preserved as evidence rather than forced into a catalog mapping.