Algorithm Pipelines

A public view of core algorithmic flows used by the system.

Back to Introduction

This page documents the core algorithms and data pipelines so stakeholders can understand how inputs are processed, where decisions are made, and which modules are involved.

Address normalization & similarity

Normalizes user-provided addresses and computes similarity for deduplication/matching.

Prefer libpostal results; fallback exists for offline/dev.
No results yet. Run the pipeline to see stage outputs.

Android attestation verification

Verifies Android SafetyNet / Play Integrity attestation tokens and extracts auth claims.

JWKS mode supports key rotation; caching is applied to reduce network calls.
No results yet. Run the pipeline to see stage outputs.

AAVA issuance pipeline

Validates consent artifacts, fetches AIA registry keys, and issues a signed AVA.

Passed policies and model versions are included in AVA metadata.
No results yet. Run the pipeline to see stage outputs.

Imagery hashing & model prediction

Computes perceptual hash and optionally calls external model for classifications.

Model is stubbed for deterministic predictions during CI.
No results yet. Run the pipeline to see stage outputs.

Authority connectors and verification

Calls external authority endpoints using HMAC or bearer tokens, with retries and backoff.

Admin UI stores vault_secret_path references instead of raw tokens for security.
No results yet. Run the pipeline to see stage outputs.

Key management & rotation

Abstracts multiple KMS providers and supports rotation and Vault integration.

Vault KV v2 is used to store credential references; system encourages referencing secrets rather than storing tokens.
No results yet. Run the pipeline to see stage outputs.