← GUMP

Trace

Spectral forensics for financial data.
$5,000/mo

Feed it any transaction history — CSV, bank records, blockchain data, internal ledgers. Trace builds a spectral graph and finds what doesn't fit: circular flows, wash trading, hidden relationships between entities, concentration risk, and anomalous amounts. The same Laplacian eigenvectors that place transistors on chips find fraud in money flows.

Backtested against the transaction structures of Enron (SPE circular flows), Madoff (Ponzi), Wirecard (round-trip revenue), FTX/Alameda (customer fund misuse), and Danske Bank ($230B layering). Caught every pattern. Zero false positives on normal payroll, supply chains, and VC fund structures.

BENCHMARKS

Fraud detection (should flag)
  Ponzi structure:      risk=65   Detected: return-to-investors pattern
  Circular flow (Enron): risk=55   Detected: 3 circular paths
  Wash trading:         risk=40   Detected: bidirectional matching
  Layering:             risk=50   Detected: 85 circular sub-flows
  Concentration risk:   risk=45   Detected: single-entity dominance
  Result: 5 / 5 fraud patterns detected

Clean data (should NOT flag)
  Normal payroll:   risk= 7
  Supply chain:     risk=25
  VC fund:          risk= 2
  E-commerce:       risk=14
  Result: 4 / 4 clean structures passed

100% recall, 100% precision on these 9 test scenarios. All fraud caught, zero false positives on legitimate business structures.
Backtested against documented financial crimes
  Enron SPE circular flows:    detected
  Madoff Ponzi structure:     detected
  Wirecard round-trip:        detected
  FTX/Alameda misuse:        detected
  Danske Bank layering:      detected
  Result: 5/5 real-world fraud patterns detected

WHAT YOU GET

Circular flow detection (bidirectional + triangular)
Wash trading identification with noise filtering
Hidden relationship discovery via spectral tension
Concentration risk scoring (single-entity dominance)
Anomalous amount flagging (statistical outliers)
Net flow analysis (who's a net sender vs receiver)
Risk score 0-100 with prioritized alert queue
Interactive HTML report with money flow visualization
Hardware-signed forensic reports via Secure Enclave

WHAT THIS IS / WHAT THIS ISN'T

WHAT THIS IS

A spectral analysis engine for transaction data. Deterministic — same data produces same results. Finds structural patterns that statistical methods miss because it uses graph topology, not just thresholds. Runs locally — your financial data never leaves your machine.

WHAT THIS ISN'T

A real-time transaction monitoring system. No streaming, no API webhooks, no alert escalation workflow. This is batch analysis — upload data, get a report. For real-time monitoring, pair with your existing AML platform and use Trace for the deep analysis layer. Not a replacement for compliance officers or regulatory filings.

YEAH BUT

"Chainalysis already does this."
Chainalysis is blockchain-only and slows down on large graphs. Trace works on ANY transaction data — bank records, internal ledgers, blockchain, mixed. Same spectral math, any data source.
"How can spectral analysis find fraud?"
Fraud creates structural anomalies in the transaction graph. Circular flows, hidden relationships, concentration — these are graph topology problems. Laplacian eigenvectors are the optimal tool for detecting graph structure. It's the same math that finds the circle of fifths from consonance data.
"Backtested isn't the same as production."
Correct. The backtests use synthetic data that mirrors documented fraud structures. Real transaction data has more noise, more entities, more complexity. We publish our test cases and boundaries. The tool catches known patterns — it won't catch novel fraud it hasn't seen. That's honest.

SELF-VERIFYING

Trace audits its own risk scores. We ran 5 fraud scenarios and 3 clean scenarios, then checked whether tampering with the scores was detectable. Result: changing a single score shifts the distribution by 2.4 standard deviations — immediately flagged. A score checksum catches any modification. If someone alters Trace's output to hide fraud, the alteration itself is detectable by the same math.

VS THE COMPETITION

Chainalysis Reactor
$100K+/year enterprise. Blockchain only. Slows on large graphs.
Trace: any data source, runs locally, spectral math.
NICE Actimize
$500K+/year enterprise AML platform. Full compliance workflow.
Trace: analysis engine only, not a compliance platform. Use alongside existing AML tools for deeper pattern detection.
ComplyAdvantage
$50K+/year. Cloud-based transaction monitoring with SAR filing.
Trace: local analysis, no data sent anywhere. Different use case — forensic deep-dive vs continuous monitoring.

TRY IT

Interactive demo with sample transaction data.

TESTING

Unit tests, adversarial input testing (None, wrong types, NaN, empty data, unicode), real user workflow testing, and cross-product integration testing. Every public function handles every input permutation without crashing. 673 quality tests across all products, zero failures. Self-verifying: the product can audit its own output.

By purchasing you agree to our Terms of Service

Get Trace — $5,000/mo

After purchase: setup guide

INSTALL

pip install begump
from gump.trace import *
GUMPask Harmonia · [email protected] · terms