🛡️ FinSentry — Financial Transaction Anomaly & AML Detection

Alert queue — accounts at MEDIUM risk or above

Transaction timeline

Money-flow network — top-risk accounts & counterparties

Circular money flows detected

Ensemble model score distributions

What this is

FinSentry scores every account in a transaction ledger for money-laundering risk by combining three independent signals — exactly the pattern real transaction-monitoring platforms use:

• AML typology rule engine — structuring, smurfing, rapid movement / layering, circular flows (round-tripping), high velocity, cash-intensive, cross-border to high-risk jurisdictions, round amounts, odd-hour activity. Each fires a weighted, human-readable reason code.
• Unsupervised ML ensemble — Isolation Forest + Local Outlier Factor + robust Mahalanobis distance over standardized behavioural features.
• Money-flow graph analysis — circular flows, fan-in/out hubs and betweenness over material (high-value) flows.

The three are blended into a 0–100 risk score, banded LOW / MEDIUM / HIGH / CRITICAL, and ranked into an alert queue with reason codes. All data here is synthetic — generated with reproducible seeds, no real customer information.

This page is a static snapshot of a real pipeline run. The fully interactive tool (adjust data size, weights, upload your own ledger) is the Streamlit app in the GitHub repo.