Skip to main content
Back to challenges
Challenge

Orchestrate Multi-Agent Real-Time Payment Fraud Triaging with AutoGen

Instant payment rails require sub-second fraud detection and automated investigator queue management. Build an AutoGen multi-agent system where specialized analyst agents collaborate to triage high-speed ISO 20022 messages, reducing investigation latency by 75%.

CybersecurityHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Build an AutoGen multi-agent system of specialized investigators to inspect and triage instant payment transactions.

Delivery guide

How work is evaluated

Evaluation

Evaluates the triage accuracy and speed of AutoGen agent conversations on ISO 20022 streams.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

High Value Offshore Hold Test

High value transactions to offshore entities must be held.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

Triage Precision

Accuracy of suspicious transaction flags. • target: 0.96 • range: 0.9-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Configure an AutoGen team consisting of a Transaction Parser, Graph Analyst, and Triage Lead.

  • Parse and validate structured ISO 20022 XML/JSON payment payloads in real time.

  • Facilitate automated consensus discussions among agents regarding transaction hold status.

  • Reduce manual investigator queue backlog while suppressing false alarms.

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic stream of ISO 20022 pacs.008 messages containing embedded fraud signals.

Challenge setup

Evaluates the triage accuracy and speed of AutoGen agent conversations on ISO 20022 streams.

Challenge input

XML/JSON ISO 20022 payment string (pacs.008)

AutoGen

Enables role-playing agent conversations for risk consensus.

Evaluated output

JSON with decision ('PASS', 'HOLD', 'BLOCK'), consensus_confidence

Evaluation checks
  • High value transactions to offshore entities must be held.
  • Accuracy of suspicious transaction flags. • target: 0.96 • range: 0.9-1
Configured evaluation evidence
  • Benchmark: ISO20022FraudBench
  • Triage Precision target: 0.96
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • AutoGen · Required
  • Microsoft AutoGen · Optional

Evaluation contract

  • High Value Offshore Hold Test · Weight 1
  • Triage Precision · Weight 1

Metrics

  • Triage Precision · Target 0.96

Benchmarks

  • ISO20022FraudBench
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: XML/JSON ISO 20022 payment string (pacs.008)

    Output: JSON with decision ('PASS', 'HOLD', 'BLOCK'), consensus_confidence

    Execution harness: Python

Evaluation tools

  • AutoGen · Agent Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Orchestrate Multi-Agent Real-Time Payment Fraud Triaging with AutoGen