Financial Mule Detection

Money mules are individuals who transfer illicit funds on behalf of criminals, either knowingly or unknowingly. They are instrumental in laundering money obtained from online scams and phishing activities.

Over 4.5 lakh mule accounts were frozen in the past year alone, with cybercrime losses estimated at ₹17,000 crore. Mules keep evolving their behaviour to avoid detection, which makes rule-based algorithms redundant as they produce plenty of false positives.

The Centre is collaborating with one of the largest private lenders in the country to build a deep learning based model to detect mules in real-time.

This problem is complex primarily due to three reasons:-

• Severe class imbalance (< 0.0035 % True Positives),
• The business impact of false positives is disproportionately higher.
• Noisy and imprecise labels.

We exploit differences between mule and regular accounts in terms of transaction patterns, demographics, and bureau information. The goal is to identify hidden patterns and outliers that indicate anomalous behaviour. The solution combines unsupervised and supervised techniques to flag fraudulent transactions.