Gajendra Deshpande - Fighting Money Laundering with Python and Open Source Software | PyData Global

Money laundering typically occurs in three stages: placement (breaking large amounts into smaller transactions), layering (moving money through multiple accounts), and integration (final transfer to receiver)

Common money laundering methods include:

• Trade-based laundering using fake/inflated invoices
• Shell companies that exist only on paper
• Real estate purchases with illicit money
• Cryptocurrency transactions
• Offshore accounts in less regulated jurisdictions
• Gambling and casino schemes

Benford’s Law is used to detect potential money laundering by analyzing the distribution of leading digits in financial data, with digit 1 occurring ~30% of the time naturally

Python tools for anti-money laundering analysis:

• NetworkX library for visualizing transaction relationships
• Pandas for data processing
• Logging module for maintaining evidence trail
• Hash functions (SHA-256, MD5) for data integrity
• Benford’s Law libraries for statistical analysis

Legal/forensic requirements:

• Must follow proper evidence collection and preservation procedures
• Analysis methods must be reproducible and peer-reviewed
• Results must be explainable in non-technical terms
• Software must not modify original evidence
• All steps must be documented for court admissibility

Politically Exposed Persons (PEPs), their relatives and close associates require extra scrutiny in financial transactions

Graph visualization is used to show transaction patterns between accounts:

• Darker arrows indicate higher frequency/amounts
• Red arrows typically indicate suspicious transactions
• Green arrows show legitimate transactions

Machine learning algorithms can be applied to transaction graphs to predict potential money laundering patterns

Open source tools may better meet legal evidence standards compared to closed source tools due to transparency and peer review capabilities