Good First Research Problems

Good first PET research problems should produce a reusable artifact: a checklist, benchmark, reproduction, small tool, failure-case catalog, or measurement report.

Rewrite PET Claims With Threat Models

Field	Card
Problem	Can common PET marketing claims be rewritten as precise threat-model statements?
The itch	Readers see phrases like "privacy-preserving AI" without knowing what is protected or from whom.
Why it matters	Bad claims lead to bad procurement, weak reviews, and overtrust.
Current workaround	Experts mentally translate claims during review.
Why the workaround is insufficient	The translation is not reusable and non-experts do not see it.
What good progress would look like	A public set of before/after claim rewrites with protected assets, adversaries, assumptions, and residual risks.
Difficulty	Good first research problem
Good for	Privacy engineer, policy researcher, technical writer
Related PETs	All PETs
Possible first contribution	Rewrite ten PET claims from public materials and publish a scoring rubric.

Small-Cohort Aggregate Leakage Catalog

Field	Card
Problem	What aggregate outputs leak sensitive facts when cohorts are small?
The itch	Teams often assume aggregates are safe until a rare group, hospital, advertiser, or region is isolated.
Why it matters	Small-cohort leakage affects healthcare, public statistics, advertising, and fraud analysis.
Current workaround	Use rough minimum thresholds.
Why the workaround is insufficient	Thresholds are often arbitrary and do not explain residual leakage.
What good progress would look like	A catalog of small-cohort leakage examples with mitigations such as grouping, suppression, DP, and output review.
Difficulty	Good first research problem
Good for	Privacy engineer, benchmark maintainer, policy researcher
Related PETs	DP, federated analytics, clean rooms, MPC
Possible first contribution	Build five synthetic examples where aggregates reveal an individual or organization and test common thresholds.

PET Setup Friction Diary

Field	Card
Problem	Where do developers get stuck when trying PET tools for the first time?
The itch	Tooling pain is often discussed anecdotally rather than measured.
Why it matters	Adoption fails when setup, docs, and examples are too hard for the target user.
Current workaround	Maintainers wait for bug reports or support tickets.
Why the workaround is insufficient	It captures only users who persist long enough to complain.
What good progress would look like	A reproducible setup diary that records time, errors, missing docs, security assumptions, and first successful run.
Difficulty	Good first research problem
Good for	Systems builder, benchmark maintainer, technical writer
Related PETs	MPC, HE, FL, TEEs
Possible first contribution	Try one PET tool from a clean environment and publish a timed friction report with suggested fixes.

Privacy Test For A Toy RAG System

Field	Card
Problem	What simple tests catch privacy leaks in a toy RAG system?
The itch	RAG demos often check answer quality but not unauthorized retrieval, prompt logging, or answer leakage.
Why it matters	Private RAG is a common enterprise pitch and an easy place to leak sensitive documents.
Current workaround	Manual review of prompts and outputs.
Why the workaround is insufficient	It is not repeatable and misses regressions.
What good progress would look like	A small test harness with role-based documents, adversarial prompts, expected-deny cases, and log checks.
Difficulty	Good first research problem
Good for	Systems builder, privacy engineer, benchmark maintainer
Related PETs	TEEs, confidential RAG, access control
Possible first contribution	Create a 20-document RAG fixture with three roles and tests for unauthorized snippet exposure.