Detecting interactions outside the committed evidence boundary

Determine how to detect transactions prevented from reaching the evidence witness, as well as false but internally consistent ground-truth profiles, beyond the observability provided by the ClaimReceipt system’s committed-ingress model.

Background

ClaimReceipt distinguishes several layers of verification: transport integrity, semantic replay, and prospective coverage after an experiment’s ingress has been committed. The paper identifies a residual L4 boundary beyond these mechanisms. A receipt can expose omissions among transactions that were committed to the system, but it cannot establish that an interaction occurred if the platform prevented it from reaching the witness, nor can it validate a ground-truth profile that is false while remaining internally consistent with the retained evidence.

This unresolved boundary limits the meaning of a successful coverage result: coverage is established only relative to the committed ingress universe. Addressing the problem would require mechanisms capable of witnessing or otherwise validating interactions before they can be excluded from the committed evidence set, together with methods for establishing the external truth of sensitive profile data.