HomeSelected engagements

04 / 05Case study

Graph‑based fraud ring detection

Kuda Technologies · African neobank, 7M+ customers

Organised fraud that single-account checks missed, found by mapping how accounts connect.

20+

fraud rings identified and disrupted

Fraud detection · Kuda Technologies

20+ rings

identified and disrupted, 3+ connected accounts each

Account by account each one looks ordinary
As a transaction graph the rings stand out

Illustration of the method, not real accounts. A capability the company did not have before; output fed straight into risk decisioning, triggering restrictions and recovery.

Built with NetworkX, Graph analysis

01

Problem and solution.

The problem

Organised fraud rings operating across multiple connected accounts were evading individual-account detection methods.

What I built

Initiated and built a graph-based fraud detection capability that did not previously exist at the company. NetworkX-based analysis mapping transaction relationships and behavioural patterns to surface organised multi-party fraud networks.

  • NetworkX
  • Python
  • Graph analysis
  • SQL

02

Architecture.

Schematic

  1. 01 Transaction data accounts and transactions · SQL
  2. 02 Relationship graph NetworkX · transaction relationships, behavioural patterns
  3. 03 Fraud ring detection (detection step) 3+ connected accounts
  4. 04 Risk decisioning
    • Account restrictions
    • Recovery actions
Schematic. Component names from the engagement summary.

03

Results.

  • 20+

    fraud rings identified and disrupted

  • 3+

    connected accounts in each ring

  • Output fed directly into risk decisioning, triggering account restrictions and recovery actions across affected networks.