Reader Paths

This page gives a short, ordered reading sequence for five common intents. Follow the path that matches your immediate goal, then stop when you can write a specific decision. Reading more pages is not progress if the protected asset, adversary, allowed output, and benchmark are still vague.

I need to choose a PET

You have a real system, a data constraint, and a privacy requirement. You need a shortlist of candidates with reasons to accept or reject each.

What Are PETs? — one-page orientation on what PETs do and do not guarantee.
PET Taxonomy — families, properties, and relationships between PETs.
Choose a PET — the primary guided chooser; answers depend on data movement and adversary.
By Data Movement — filter by where data lives and moves.
By Threat Model — filter by who must not learn what.
Decision Tree — a readable branching checklist to confirm or challenge your shortlist.

Exit when you can state: recommended PET, credible alternatives, rejected PETs, the assumption that matters most, and the measurement that would change the choice.

I need to design an architecture

You have a PET candidate and need to produce a defensible system design with actors, data flows, trust boundaries, and failure modes.

What Are PETs? — confirm your candidate fits the protection goal.
PET Architectures: Overview — the catalogue of reference architectures.
Pick the architecture closest to your design (for example, FL + Secure Aggregation, MPC Analytics Pipeline, or Confidential RAG).
Threat Models: Overview — map your adversary assumptions to the standard models.
Benchmarks: Overview — decide what to measure before you build.

Exit when every boundary in the diagram has an owner, an artifact crossing it, and a failure mode. A PET architecture without a logging and output story is only half designed.

I need to evaluate a claim

A paper, vendor, or colleague is making a privacy claim. You need to decide whether it is credible and what it actually protects.

Glossary — resolve any terms you are uncertain about.
Evidence Policy — the standard of evidence this guide uses; apply the same bar to external claims.
Claim Register — the live register of tracked claims; check whether the claim has already been assessed.
Threat Models: Overview — identify the adversary the claim assumes and what it does not cover.
By Threat Model — check whether the chosen PET matches the stated threat.

Exit when the claim can be rewritten as: this system protects a named artifact from a named adversary under named assumptions while revealing a named output.

I need a research problem

You want a concrete open problem with a clear gap, current workaround, success criteria, and a plausible first contribution.

Fix My Itch: Overview — the full index of open problems by PET family.
Good First Research Problems — entry-level problems with lower coordination cost.
Industry Pain Points — problems that practitioners have explicitly flagged as blockers.
Benchmarks Needed — measurement gaps where a new benchmark would have high impact.
Browse the PET-specific problem lists (e.g., Federated Learning, Differential Privacy, MPC) to find the family that interests you.

Exit when you can name who would use the result, why the current workaround is insufficient, and what a first reproducible artifact would look like.

I need evidence from deployments

You want to understand how a PET has performed in production, what the maturity level is, and what caveats remain unresolved.

Evidence Policy — understand how deployment evidence is graded before reading the entries.
Deployments: Overview — the full index of deployment write-ups, filtered by domain.
Browse by domain: Healthcare, Finance, Advertising, Public Sector.
Claim Register — cross-reference specific deployment claims that are under review.
Benchmarks: Scorecards — compare deployments on privacy, utility, and cost dimensions.

Exit when you can separate production evidence from pilots, vendor case studies, illustrative examples, and expert judgment.