Detect dense UPI/device/IP sharing networks in real time. This “RingDetect” demo hits TigerGraph REST++ through a tokenless proxy so judges can see the signal without exposing secrets.
Shared devices, phones, or IPs hint at coordinated mule activity. Without relationships, rules flood analysts with false positives. RingDetect links the graph, surfaces the influencers, and lets you rerun with new thresholds instantly.
Your proxy (Flask + Gunicorn) handles CORS and forwards `/api/ringDetect` to TigerGraph. In local mode it talks to `http://localhost:9000` without any Savanna token; if you deploy the proxy remotely, a bearer token is optional.
Investigators can focus on accounts tied to high-risk merchants or shared devices. RingDetect explains why each account is suspicious (risk_score, shared_count, degrees), making it easy to tell the story during demos or in production.
Legacy systems treat transactions in isolation. Coordinated rings slip through when shared identifiers are ignored or too costly to compute over wide datasets.
Fraud Graph links accounts to devices, merchants, phones, and IPs. Queries like RingDetect and MuleRanking expose the clusters and rank suspicious merchants so an analyst can validate the ring in minutes.
Less manual triage, faster investigations, and auditable insights for compliance teams. Prove the concept with the live viewer locally first, then swap in a cloud proxy only if you want to deploy it remotely.
Explore the ring in TigerGraph-style view: the source account sits in the center, suspects are arranged around it, and links show how strongly each account connects back to the ring.