How I Caught a Massive Layering Scheme in Mobile Banking

Photo by Roger Starnes Sr on Unsplash

I still remember the day our team detected a massive layering scheme in our mobile banking system. It was a typical Monday morning when our alert system started buzzing with unusual transaction patterns. The numbers were staggering - over 10,000 transactions in a single day, all below the BDT 100,000 threshold, and all of them were layered in a way that seemed almost impossible to detect.

The Hidden Problem

As I dug deeper, I realized that our AML rule engine was missing a critical aspect of layering detection. The engine was designed to catch obvious structuring attempts, but it was not sophisticated enough to identify complex layering schemes. This was a major concern, as layering is a common technique used by money launderers to evade detection.

According to the BFIU guidelines, layering is defined as the process of moving funds through multiple transactions to disguise the origin of the money. In mobile banking, layering can be particularly challenging to detect, as transactions are often small and frequent. The BDT 100,000 threshold monitoring is in place to prevent structuring, but it is not enough to catch layering schemes.

Technical Breakdown & Logic Flow

To detect layering, we needed to analyze transaction patterns and identify unusual behavior. We started by collecting data on all transactions below the BDT 100,000 threshold. We then applied a series of filters to remove legitimate transactions, such as transactions between known merchants and customers. The remaining transactions were then analyzed for layering patterns.

The logic flow was as follows:

1. Collect transaction data
2. Apply filters to remove legitimate transactions
3. Analyze remaining transactions for layering patterns

We used a combination of machine learning algorithms and rule-based systems to detect layering patterns. The machine learning algorithms were trained on historical data to identify unusual transaction patterns, while the rule-based systems were designed to catch specific layering techniques.

Python Implementation

import pandas as pd
from sklearn.ensemble import IsolationForest

# Load transaction data
transactions = pd.read_csv('transactions.csv')

# Apply filters to remove legitimate transactions
transactions = transactions[transactions['amount'] < 100000]
transactions = transactions[transactions['type'] == 'cash_in']

# Analyze remaining transactions for layering patterns
layering_patterns = transactions.groupby('customer_id')['transaction_id'].count()
layering_patterns = layering_patterns[layering_patterns > 10]

# Use machine learning algorithm to detect unusual transaction patterns
ml_model = IsolationForest(contamination=0.1)
ml_model.fit(transactions)

# Identify transactions that are likely to be part of a layering scheme
suspicious_transactions = ml_model.predict(transactions)

The code above shows how we used Python to detect layering patterns in our transaction data. We started by loading the transaction data into a Pandas dataframe. We then applied filters to remove legitimate transactions, such as transactions between known merchants and customers. The remaining transactions were then analyzed for layering patterns using a combination of machine learning algorithms and rule-based systems.

Local Application

In Bangladesh, mobile banking is a popular way to transfer money, especially among the unbanked population. However, this has also created opportunities for money launderers to exploit the system. The BFIU guidelines require mobile banking operators to monitor transactions and report suspicious activity.

Our solution was designed to meet these requirements. We worked closely with the BFIU to ensure that our system was compliant with their guidelines. We also ensured that our system was able to detect layering schemes that were specific to the Bangladeshi market.

Common Pitfalls & Edge Cases

One of the common pitfalls we encountered was the high number of false positives. Our system was designed to detect unusual transaction patterns, but it was not perfect. We had to continuously fine-tune our algorithms to reduce the number of false positives.

Another edge case we encountered was the use of multiple accounts to layer transactions. Money launderers would often use multiple accounts to move funds, making it difficult to detect layering schemes. We had to design our system to detect these types of schemes.

Counterintuitive Insight

One of the counterintuitive insights we gained from our experience was that layering schemes are often more complex than they seem. Money launderers will often use multiple layers to disguise the origin of the money, making it difficult to detect.

However, this complexity also creates opportunities for detection. By analyzing transaction patterns and identifying unusual behavior, we can detect layering schemes that might otherwise go undetected.

Conclusion & CTA

In conclusion, detecting layering schemes in mobile banking requires a combination of machine learning algorithms and rule-based systems. Our experience has shown that it is possible to detect layering schemes, even in complex cases.

So, what's the weirdest transaction pattern you've seen? Have you encountered any layering schemes in your experience? Drop a comment below and let's discuss. Also, check out our other resources on aitipseveryday.com for more information on AML and compliance.