Trustra
Trust scoring engine with fraud detection, network influence propagation, and abuse simulation.
Overview
Trustra computes dynamic trust scores for users in a peer-to-peer transaction network. Scores are derived from 6 weighted signals, updated on a schedule, and explained per recalculation. An abuse detection layer flags suspicious behavior patterns in real time. A simulation layer lets you trigger malicious scenarios on demand to observe how the scoring system responds.
The interesting part of Trustra is not just the score itself but the auditability layer underneath it. Every recalculation stores a human-readable explanation row, weights are configurable via an admin API without redeployment, and the simulation endpoints make it possible to stress-test the detection logic against realistic attack patterns.
Stack
Highlights
- Computes trust scores from 6 weighted signals: success rate, dispute rate, feedback rating, network trust influence, inactivity decay, and abuse penalties.
- Detects abuse via 3 rules - transaction spike, fake feedback burst, and cluster behavior - with flags surfaced via API and dashboard.
- Stores a per-recalculation explanation trail and full score history for time-series auditing.
- Ships simulation endpoints to trigger normal, malicious cluster, and spike scenarios on demand for demo and stress-testing flows.
