Fraud Detection & Trust Score
Understand how Begini identifies suspicious behaviour and measures the reliability of each session.
Overview
Alongside the risk score, Begini evaluates the reliability of each assessment session through a Trust Score. This is separate from the risk score — it does not indicate whether a user is high or low risk, it indicates how much confidence to place in the result.
What the Trust Score answers
"How reliable is the behavioural data from this session?"
A low trust score means the data quality is insufficient to rely on for automated decisioning, not that the user is fraudulent. The appropriate response is review, not automatic decline.
What drives a low Trust Score
| Signal | What it may indicate |
|---|---|
| Extremely fast or uniform responses | Random answering or scripted behaviour |
| Inconsistent patterns across tasks | Low engagement or external assistance |
| Repeated attempts by the same UID | Attempts to game the system |
| Behaviour inconsistent with natural interaction | Coached or assisted completion |
| Significant delays during tasks | External research during the session |
| High drop-off or irregular session flow | Disengagement or interruptions |
These signals are captured passively during the assessment — no additional input from the user is required.
How to use Trust Score alongside Risk Score
| Risk | Trust | Recommended action |
|---|---|---|
| Low | High | Approve with confidence |
| Low | Low | Review before approving |
| High | High | Decline or refer per policy |
| High | Low | Review — may be unreliable signal |
Define internal rules for how your team handles low-trust sessions. At minimum, low-trust results should not be used for fully automated decisions without a manual review step.
Detecting fraud and manipulation
Begini's behavioural approach identifies patterns that are difficult to replicate deliberately:
- Randomised answering produces unnatural timing and consistency signatures
- Scripted or coached responses create cross-task inconsistencies
- Repeated attempts to improve scores leave behavioural traces
These signals are surfaced through the Trust Score rather than requiring separate fraud tooling.
Next steps
- Understanding Assessment Results — how to use risk and trust scores together in decisioning
- Assessment Troubleshooting — diagnose issues with session quality and results
- AI-Resistant Assessment Design — how Begini makes AI-assisted completion difficult and detectable
- KYC and Identity Binding — how Begini links the assessment to a verified loan applicant
Was this article helpful?
Give feedback