☀️🌙 LIGHT
All engines ready
📊
Drop your file here
Upload bank statements, invoices, transaction records, receipts,
or any financial spreadsheet for deep fraud analysis
.xlsx .xls .csv .tsv
📐 Benford's Law
Leading digit frequency analysis — natural numbers follow predictable distributions; fabricated data doesn't
🔁 Duplicate Detection
Exact & near-duplicate transactions, same amount same vendor, split payments below thresholds
📈 Statistical Outliers
Z-score, IQR, and modified Z-score methods to flag statistically abnormal amounts
🎯 Round Number Bias
Disproportionate round numbers ($100, $500, $1000) are a common fabrication indicator
🕐 Temporal Patterns
Weekend/holiday transactions, velocity spikes, unusual hours, end-of-period clustering
🤖 ML Autoencoder
TensorFlow.js neural network trained on your data to detect multi-dimensional anomalies
⚡ Velocity Analysis
Transaction frequency spikes, same-vendor clustering, rapid sequential transactions
🔢 Sequence & Gap Analysis
Missing invoice numbers, non-sequential IDs, gaps that suggest deleted records
💰 Threshold Hunting
Amounts just below approval limits ($999, $4999) — classic structuring/smurfing pattern
Map your columns
FraudLens detected the following columns. Map them to analysis fields — the more you map, the more checks we can run.
TensorFlow.js ML engine will train on your data
Running analysis...
--
RISK SCORE

Analyzing...

Benford's Law — First Digit Distribution
Expected vs actual frequency of leading digits. Significant deviation suggests fabricated or manipulated data.
Expected (Benford) Actual (Your Data)
Chi-Square Test Result
Statistical significance of deviation from Benford's distribution
Mean Absolute Deviation
Average deviation per digit from expected Benford frequency
Amount Distribution
Histogram of transaction amounts — spikes at round numbers are suspicious
Threshold Proximity Analysis
Transactions suspiciously close to common approval thresholds
Vendor/Payee Concentration
Top vendors by total spend — concentration risk and outliers
Round Number Distribution
Percentage of transactions ending in .00, 000, etc.
Transactions by Day of Week
Weekends and holidays should have fewer legitimate transactions
Daily Volume Over Time
Transaction velocity — spikes may indicate burst fraud activity
Monthly Pattern
End-of-period clustering is a common manipulation indicator
Amount by Day of Week
Average transaction value per weekday — anomalous weekend amounts
TensorFlow.js Autoencoder — Reconstruction Error
Higher reconstruction error = more anomalous. The model learned your data's normal patterns and flags deviations.
Top Anomalous Transactions
Model Summary