Privacy-Preserving Authentication Audit Data Platform

Privacy Preserving Authentication Audit Data Platform

A governed authentication-event ingestion platform: queue intake, contract validation, HMAC tokenization of direct identifiers, curated persistence, a quarantine path for rejected records, and batch-level audit evidence. Every number and preview on this page is produced by the running service.

●  Source code CI status How to verify
Live container deployed

This page is backed by the same transformation module used by the ingestion worker. The sample records below are loaded from repository data, transformed by Python at request time, and displayed through API-backed previews.

Architecture

Source contract to audited warehouse scroll to zoom · drag to pan
SOURCES INGESTION VALIDATE AND TOKENIZE CURATED POSTGRES SERVING RBA login data set Kaggle / Zenodo, CC BY 4.0 Synthetic fixture deterministic fallback SQS queue intake LocalStack, batched receive LoginEvent contract JSON Schema 2020-12 Validation required fields, domain, version HMAC-SHA256 tokenization secret-keyed, deterministic Quarantine rejected payload plus reason user_logins curated fact, event_id primary key quarantine_login_events investigable rejects ingestion_audit received / loaded / rejected vw_ingestion_health FastAPI surface previews, schema, /healthz Docker and Caddy TLS edge, this page invalid ON CONFLICT DO NOTHING makes replay safe

Source Registry

The platform is now wired for a real authentication research dataset. The running page uses the local fixture until the RBA dataset is downloaded and normalized, then switches to the normalized RBA sample automatically.

Live Transform Preview

The source sample includes direct identifiers. The curated output replaces sensitive fields with deterministic HMAC tokens and records the source payload hash for replay and audit alignment.

Source Payload

Waiting for source sample...

Curated Output

Waiting for transform...

Data Tables

The browser shows compact 10-row previews only. The full 1.1 GB RBA source dataset is linked in Source Registry and is not hosted inside this page.

Schema and Source Contract

PostgreSQL Schema

Loading schema...

Login Event Contract

Loading contract...

Quality and Privacy Gates

These controls are implemented in code and mapped to explicit failure behavior. Invalid records are not silently corrected; they are separated for investigation and replay governance.

Operations

How work is triggered, what the last audited run did, how each failure class is handled, and what the measured numbers imply about scale and cost. The full-volume run is a point-in-time benchmark against the 1.1 GB source archive, which is not committed; correctness is verified continuously by the suite instead.

Scheduling

Scale and cost

Failure handling

Each row is a failure the platform handles today, with the code path that handles it and how far the damage spreads.

Column Lineage

Engineering Toolchain

Every tool listed is in the running system. The roadmap lives in the repository README, not on this page.

Verify It Yourself

Nothing here is a screenshot. Clone the repository and reproduce it, or call the running service directly.

# run the suite that gates every push
git clone https://github.com/svaddhiparthy/Privacy-Preserving-Authentication-Audit-Data-Platform.git
cd Privacy-Preserving-Authentication-Audit-Data-Platform
pip install -e ".[dev]" && pytest && ruff check .

# or call this deployment
curl -s https://srivaddhiparthy.com/privacy-preserving-authentication-audit-data-platform/healthz
Repository Design notes and rationale Architecture write-up