Data Anonymization Knowledge Base

An internal reference for Aliz data engineers on data anonymization and privacy engineering. Use this to ramp up before a client engagement, as a reference during implementation, or as LLM context for generating client deliverables.

How to use this knowledge base

"I'm new to anonymization" — Read the Anonymization Essentials. It teaches the core concepts through real-world failure cases — no prior knowledge needed, ~2 hours.

"I know the basics, give me the full toolkit" — Read the Crash Course. It condenses all 13 learning path modules into a single dense reference covering techniques, architecture, GCP tooling, key management, verticals, and client engagement.

"I want the deep treatment" — Read the learning path in chapter order. The Core tier in each module gives you the essentials in ~30 minutes; go deeper where your project needs it.

"I need a specific answer" — Jump to the relevant reference module or cheat sheet. These are structured for lookup, not sequential reading.

"I'm prepping a client deliverable" — Start with 13-client-engagement.md for engagement patterns and LLM-assisted drafting workflow, then grab the relevant ready-to-adapt template from templates/.

Learning path

Organized into six chapters. Each module is self-contained — terms are defined inline, no required reading order beyond the chapter grouping. Sections within each module are labeled Core (read this) and Deep dive (read when you need it).

Ch 1 — Fundamentals

Module	What it covers
01-why-this-matters	Business case, regulatory landscape, anonymization vs pseudonymization
02-foundations	Identifier types, the anonymization spectrum, threat models, measuring privacy

Ch 2 — Techniques & Models

Module	What it covers
03-techniques	Suppression, generalization, hashing, tokenization, FPE, masking, perturbation, synthetic data
04-privacy-models	k-anonymity, l-diversity, t-closeness, differential privacy

Ch 3 — Implementation

Module	What it covers
05-tooling-landscape	GCP Sensitive Data Protection, BigQuery security, Cloud KMS, Dataflow, OSS tools
06-architecture-patterns	Multi-tier data zones, pipeline-integrated anonymization, query-time masking, telco CDR and e-commerce reference architectures
07-key-management	Transform decision framework, token vaults, Cloud KMS with HSM, key rotation, crypto-shredding for right to erasure
08-ai-and-llm-patterns	ML vs LLM impact, free-text PII detection, selective masking, Model Armor, Gemini Enterprise, BigQuery Conversational Analytics

Ch 4 — Governance & Operations

Module	What it covers
09-governance-and-ops	Anonymization policy, data ownership (DAMA DMBOK), RACI, org readiness, lawful processing without anonymization, decision framework, scaling down for SMEs
10-operations-and-incidents	Automated classification (SDP discovery), pipeline monitoring, lineage and audit trails, key lifecycle, incident response, breach notification

Ch 5 — In the Field

Module	What it covers
11-industry-deep-dives	Telco, retail/e-commerce, financial services, energy — vertical-specific challenges and patterns (5 sub-modules)
12-risks-and-failures	Attack taxonomy, re-identification cases, implementation mistakes, testing and validation

Ch 6 — Client Engagement

Module	What it covers
13-client-engagement	Discovery, common scenarios, business case building, architecture recommendations, deliverable patterns

Reference modules

Lookup-friendly counterparts to the learning path — tables, reference cards, and checklists instead of narrative. Each module includes quick-navigation aids (flowcharts, signal-matching tables, or section indexes) to help you jump to the right section. Focused on delivery and presales usage.

Ordered by engagement lifecycle: presales → context → design → validate → operate.

Phase	Module	What it covers
Presales	Client Engagement	Scenario playbook, discovery checklist, business case, phased roadmap, deliverable mapping
Context	Industry Reference	Cross-vertical comparison, recurring data types, transferable architecture decisions
	— Telco	Identifier treatment, CDR cookbook, ePrivacy pipeline implications
	— Retail	Multi-domain identity, PCI-DSS tokenization, consent management
	— Financial Services	What can't be anonymized, regulatory collisions, fraud detection
	— Energy	NILM inference, smart meter treatment, metering regulation
Design	Techniques	Per-technique cards with mechanism, reversibility, GCP support, risks + comparison matrix
Design	Privacy Models	Model specifications, parameter guidance, selection guide
Design	Architecture Patterns	Pattern catalog + key management + AI/LLM patterns
Design	Tooling	GCP services (SDP, BigQuery, KMS) + OSS tools (ARX, OpenDP, Presidio) + configs + pricing
Validate	Risks and Failures	Attack catalog, case studies, implementation mistakes diagnostic checklist
Operate	Governance and Ops	Checklists, RACI, monitoring signals, incident response, audit readiness

Cheat sheets

One-page quick references — designed for pinning to a wall or pulling up mid-conversation.

Sheet	What it covers
Engagement Checklist	Print before a client meeting — 5 discovery questions, scenario signals, quick wins, deliverable mapping
Zone Architecture	The whiteboard diagram — 4-zone model, vertical adaptation, 3 transferable architecture decisions
Attack–Countermeasure Mapping	Attack → countermeasure → GCP tool lookup + "is it really anonymized?" quick test

Templates

Resource	What it covers
one-pager-technique	Ready-to-adapt technique document (example: HMAC-SHA256 for MSISDN pseudonymization)
one-pager-regulatory	Ready-to-adapt regulatory briefing (example: GDPR anonymization requirements)
one-pager-architecture	Ready-to-adapt architecture recommendation (example: four-zone telco CDR on GCP)
anonymization-policy	Ready-to-adapt anonymization policy with pre-populated example content

Other resources

Resource	What it covers
glossary	Term definitions compiled from all modules