Mule account intelligence
Accounts flagged by the behavioural engine as probable mules, scored before any registry match. This is the module that feeds the bank's own contributions into the Smart Registry rather than only consuming from it.
Detection queueranked by mule probability
| Account | Mule probability | Pattern | Flag age | Credits | EDD status | Registry | |
|---|---|---|---|---|---|---|---|
| 50100777333221 | 94% | Fan-in 23 → single withdrawal | 6 days | ₹4.2L / 23 txn | EDD in progress | Not yet reported | Report → |
| 50100445566778 | 88% | Dormant 14mo → burst activity | 3 days | ₹2.8L / 11 txn | EDD in progress | Not yet reported | Report → |
| 50100889900112 | 81% | Device shared with SR-4102 | 9 days | ₹1.1L / 7 txn | Awaiting EDD | Not yet reported | Report → |
| 50100223344556 | 76% | Structuring — ₹49,500 x 6 | 2 days | ₹2.97L / 6 txn | Awaiting EDD | Not yet reported | Report → |
| 50100999888777 | 72% | Velocity — 14 new beneficiaries | 1 day | ₹90k / 14 txn | Reported as ALERT | SR-4103 | Report → |
| 50100112233445 | 58% | Salary account, single spike | 12 days | ₹65k / 2 txn | Cleared on EDD | Not reported | Report → |
How a mule gets promoted. Behavioural detection produces a candidate. Only after Enhanced Due Diligence confirms the behaviour does the status move from Alert to Suspect in the registry. An unverified model output must never be published as Suspect — that is a customer-harm event and a DPDP fairness exposure.
Pattern librarybehavioural engine
Fan-in / fan-out
Many small credits converging, then one outbound sweep.
Dormancy then burst
A long-idle account waking with high velocity.
Structuring
Repeated amounts sitting just under a reporting threshold.
Device & IP sharing
Shared hardware with a reported entity — weak alone.
Profile change then transfer
Contact details swapped shortly before a payout.
Beneficiary velocity
New payees added far faster than the customer's baseline.