Tamper-proof records. Examiner-ready evidence. One API call. The cryptographic compliance layer for modern fintech.
From fraud decision to examiner-ready evidence in four steps.
Your fraud model makes an ALLOW, BLOCK, or REVIEW decision on a transaction.
Attestr hashes it, chains it to the previous record, signs it with Ed25519, and stores it in an append-only ledger.
Months later, a regulator asks about a blocked payment. Your analyst opens the investigation view.
A cryptographically verifiable evidence packet proves exactly what happened, when, and why. Tamper-proof.
Six layers of cryptographic integrity. Zero trust required.
Every record includes the previous record's SHA-256 hash. Altering any single record breaks every subsequent link.
Ed25519 cryptographic signatures on every record. Download the public key, verify any evidence packet offline.
Records are batched into Merkle trees. Prove inclusion of any single event within a batch of 1,000+ records.
PostgreSQL rules enforce immutability at the database level. No UPDATE, no DELETE. Records are permanent.
Every API request is signed with SHA-256 HMAC including method, path, timestamp, and body. 5-minute replay window.
Isolated data per tenant. Each organization gets unique API credentials, separate ledger chains, and independent verification.
From “trust us” to mathematical proof.
When an examiner asks about a blocked transaction, you don't explain — you prove. Evidence packets contain everything needed for independent verification.
Recompute the SHA-256 hash from the canonical record. If it matches, the data is unaltered.
Verify the previous_hash matches the prior record. The chain is unbroken.
Verify the Ed25519 signature using the public key. Attestr signed this record.
Walk the Merkle proof path to the root. This record was in the batch.
Compare the input_hash against the source system's raw data. The decision was based on real data.
Third parties can verify evidence packets independently using the public key endpoint. No Attestr account required.
SELF-HOST OPTION: Tenants can host their own verification portal using our open-source engine. Zero vendor lock-in.
Prove the input, not just the output.
Anyone can sign a decision. But can you prove the data that fed the decision wasn't tampered with? Input hashing closes the last trust gap.
Your fraud engine sends a decision. Attestr signs it. But an insider could have altered the transaction data before the engine saw it. You've notarized a lie.
Before your engine runs, compute a SHA-256 of the raw transaction payload. Send it alongside the decision. Now anyone can pull the original data from the source system and verify the hash matches.
The input_hash field is optional. Records without it are still fully signed and chained (single attestation). Adding it upgrades to dual attestation — binding the decision to its source data.
Decision is signed, chained, and Merkle-batched. Proves the record wasn't altered after ingestion.
All of single attestation, plus a cryptographic anchor to the original input data. Full chain of custody.
Your ledger doesn’t just record. It watches.
Statistical anomaly detection runs continuously on your decision stream. No external APIs. No data leaves your infrastructure. Pure math.
Detects when your model’s score distribution shifts. Catches model drift before examiners do.
Rolling z-score analysis on decision scores. Alerts when the current window diverges >2σ from the 24-hour baseline.
Flags unusual spikes in block/allow ratios. Know when something changes before your users call.
Proportion test comparing current-hour decision ratios against 24-hour rolling averages.
Detects missing records and ingestion failures. Proves you never stopped recording.
Monitors ingestion velocity per tenant. Alerts when throughput drops below 20% of expected rate.
Surfaces new fraud patterns and policy changes through reason code frequency analysis.
Cosine similarity between current and baseline reason code frequency vectors.
Automatically detects model deployments and correlates with decision pattern changes.
Deterministic detection of new model_version strings not seen in the 24-hour baseline.
Catches batch processing errors, replay attacks, and upstream integration failures.
Z-score analysis on ingestion rate. Flags when velocity exceeds 3σ above baseline.
All analysis runs locally on your infrastructure.
No data is sent to external AI services. Detection uses deterministic statistical methods — z-scores, proportion tests, cosine similarity. Fully auditable. No black boxes.
Paste JSON. Watch it chain.
See how a fraud decision gets hashed, chained, and signed in real time. This is a client-side demo — nothing leaves your browser.
One endpoint. Full audit trail.
Your fraud engine already makes the decision. Just POST it. We handle the hashing, chaining, signing, and Merkle batching. No SDK required.
What happens when the examiner calls.
Write SQL queries to reconstruct decision logs from scattered tables
Format CSVs, cross-reference timestamps, fill gaps in audit trail
Explain log inconsistencies to a skeptical examiner in writing
Re-pull data after examiner finds discrepancies in your first export
Pray the timestamps are close enough and the examiner doesn’t dig deeper
Plus the stress. Plus the risk. Plus the billable hours.
Examiner requests evidence for transaction txn_abc123
Open dashboard, search event_id, click “Export Evidence”
Send the PDF. Hash chain, signature, Merkle proof — all included.
Examiner verifies independently using the public key. No questions.
Go back to building.
Mathematical proof. No explaining. No stress.
Simple, predictable pricing.
Start free. Scale when you're ready. No hidden fees. No per-seat charges. You pay for records ingested.
For prototyping and evaluation.
For growing fintechs in production.
For compliance-heavy operations.
Your fraud engine already makes the right calls. Attestr makes them audit-proof. Start recording in under 5 minutes.
Open source engine · MIT Licensed · Self-host or use our cloud