Skip to main content
Back to challenges
Challenge

Detect Sanctions Evasion and High-Risk Crypto Transfers using Google ADK and Giskard

Offshore crypto transactions pose severe sanctions enforcement risks for global exchanges. Develop a Gemini-powered Google ADK agent to audit cross-border wallet histories, utilizing Giskard to stress-test AML detection models against adversarial evasion tactics.

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 AML screening agent using Google Agent Development Kit and evaluate model robustness against evasion tricks with Giskard.

Delivery guide

How work is evaluated

Evaluation

Runs Giskard model quality tests on sanctions evasion detection accuracy.

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

Sanctioned Entity Flag Test

Must flag mixing protocol interactions as HIGH risk.

Binary check

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

Dimension 2

Detection Recall

Ratio of correctly identified sanctions evasion attempts. • target: 0.98 • range: 0.95-1

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

  • Build a Google ADK agent with Gemini to inspect blockchain transaction metadata.

  • Integrate Giskard to run automated hallucination and bias scans on compliance risk reports.

  • Detect obfuscated mixing service interactions and multi-hop chain hops.

  • Generate automated regulatory Suspicious Activity Reports (SARs) with full evidence chains.

Resources and assets

Reference links and supporting material

Dataset notes

On-chain transaction graphs involving known sanctioned addresses and mixer hops.

Challenge setup

Runs Giskard model quality tests on sanctions evasion detection accuracy.

Challenge input

JSON list of wallet transfers with timestamp, amount, and hop addresses

Google ADK

Orchestrates Vertex AI Gemini models for compliance reasoning.

Giskard

Provides model risk testing and failure mode automated discovery.

Evaluated output

JSON with sanctions_risk_level ('LOW', 'MEDIUM', 'HIGH'), flag_reason

Evaluation checks
  • Must flag mixing protocol interactions as HIGH risk.
  • Ratio of correctly identified sanctions evasion attempts. • target: 0.98 • range: 0.95-1
Configured evaluation evidence
  • Benchmark: CryptoAMLBench
  • Detection Recall target: 0.98
  • 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
  • Google · Required
  • Google ADK · Optional
  • Giskard · Optional

Evaluation contract

  • Sanctioned Entity Flag Test · Weight 1
  • Detection Recall · Weight 1

Metrics

  • Detection Recall · Target 0.98

Benchmarks

  • CryptoAMLBench
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON list of wallet transfers with timestamp, amount, and hop addresses

    Output: JSON with sanctions_risk_level ('LOW', 'MEDIUM', 'HIGH'), flag_reason

    Execution harness: Python

Evaluation tools

  • Google ADK · Agent Framework
  • Giskard · AI Workflow Automation

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Detect Sanctions Evasion and High-Risk Crypto Transfers using Google ADK and Giskard