Background

Mule accounts—bank accounts opened and operated by individuals (knowingly or unknowingly) to receive, move, or withdraw proceeds of cybercrime—are increasingly used in investment frauds, phishing, instant loan app scams, ransomware, and online job scams. Cybercriminals exploit these accounts to obscure the money trail and launder illicit funds.

Despite red flags such as unusual transaction patterns, geographic inconsistencies, or short account lifespans, detecting mule accounts at scale remains a challenge due to fragmented data, lack of automation, and evolving criminal tactics.

Problem Statement

Design and build an intelligent system to identify, analyse, and visualize potential mule bank accounts involved in cybercrime. The tool should ingest transactional, KYC, and behavioural data (anonymized/simulated), apply detection logic or AI/ML models, and flag suspicious accounts for investigation with supporting evidence.

Key Objectives

• Ingest and process large volumes of bank transaction data.
• Identify red flags indicative of mule activity (e.g., rapid in-and-out transactions, cash
• withdrawals after receiving suspicious funds).
• Link connected accounts via transaction graphs to detect networks.
• Provide dashboards for LEAs or banks with risk scoring and account behaviour insights.
• Allow tagging and feedback to improve detection accuracy.

Functional Requirements

• Import structured datasets (CSV/Excel) of anonymized account transactions and KYC profiles.
• Apply rules or ML to detect mule-like behaviour.
• Generate account profiles with visual graphs showing connections, volume spikes, and flow of funds.
• Filter/search by account type, risk score, transaction size, or location.
• Export findings into investigation-ready reports.

Evaluation Criteria

• Detection accuracy (true positives/false positives).
• Innovation in network linkage or risk scoring approach.
• Data visualization and investigative usability.
• Scalability and performance with large datasets.
• UI/UX and interactivity.

Suggested Tools/Technologies

Python, Pandas, NetworkX, Scikit-learn, Neo4j, Elasticsearch, Kibana, D3.js, Flask/Django, React, PostgreSQL/MongoDB

Bonus Points

• Use of graph-based anomaly detection.
• Ability to ingest real-time transaction feeds or simulate them.
• Integration with alerting or case management system.

Deliverables

• A working prototype or web application.
• Documentation of approach, assumptions, and usage.
• Sample test cases and visual output (PDF/CSV exports, graphs, dashboards).

Tool Design Guidelines

• ACCOUNT OPENED / ACTIVATED RECENTLY
• FREQUENT PANNY TRANSACTIONS (SMALL CREDIT/DEBIT)
• TRANSACTIONS WITH BIG AMOUNT IN SHORT PERIOD OF TIME (BEHAVIORAL ANAMOLY)
• KYC MISMATCH (E. G. EMAIL ID DOES NOT MATCH WITH ACCOUNT HOLDER NAME)
• KYC ANNUAL INCOME CLAUSE MISMATCH WITH TRANSACTION AMOUNTS
• FREQUENT ADDITION / DELETION OF BENEFICIERIES FROM DIFFERENT PART OF COUNTRY
• CONSECUTIVE CREDIT AND DEBIT TRANSACTIONS WITH SAME AMOUNT VALUE
• OTHER SPECIFICATIONS