awesome-pet-stack/README.md

Awesome PETs — Privacy Enhancing Technologies

A curated map of the privacy-enhancing technology stack — the cryptographic primitives, protocols, and systems you use to build privacy into products and research.

For builders and researchers, not switchers. If you're looking for "use this app instead of Google," see awesome-privacy. This list answers the other question: which primitive, protocol, or system do I reach for?

PETs are the technologies that let data be used without being exposed — computed on, queried, shared, or transmitted while staying private. This list organizes them into a practical 7-tier stack (synthesized from the ENISA, UN, and OECD PET taxonomies) and links out to the excellent per-discipline lists for depth.

🚧 Skeleton / work-in-progress. Seeded with well-known projects; entries to be verified, expanded, and annotated. Contributions welcome.

Contents

• 1. Input Privacy — data never seen in the clear
• 2. Output Privacy — aggregate results with formal guarantees
• 3. Federated & Distributed Analytics
• 4. Anonymity & Traffic-Analysis Resistance
• 5. Verified / Proof-Based Privacy
• 6. Hardware-Anchored Confidential Compute
• 7. Accountability & Transparency
• Related Lists
• Contributing

---

1. Input Privacy

Compute on data that is never decrypted or revealed to the other party.

Secure Multi-Party Computation (MPC)

• MP-SPDZ — versatile MPC framework, many protocols.
• EMP-toolkit — efficient MPC (garbled circuits, OT).
• MPyC — Python MPC via secret sharing.

Homomorphic Encryption (HE / FHE)

• OpenFHE — community FHE library (BGV/BFV/CKKS/TFHE).
• Microsoft SEAL — widely used HE library.
• TFHE-rs / Concrete — Zama's Rust FHE stack.
• Lattigo — Go lattice-based HE.

Private Information Retrieval (PIR) / PSI / ORAM

• SealPIR — single-server PIR.
• Private Set Intersection (libPSI / volePSI) — fast PSI.

2. Output Privacy

Release aggregate insights with mathematical privacy guarantees.

• OpenDP — differential privacy primitives (Harvard/Microsoft).
• Google differential-privacy — DP libraries + tooling.
• Opacus — DP training for PyTorch.
• SDV — synthetic data generation.

3. Federated & Distributed Analytics

Train and analyze without centralizing the raw data.

• Flower — framework-agnostic federated learning.
• TensorFlow Federated — FL in TF.
• OpenFL — Intel's federated learning.
• PySyft — privacy-preserving ML (OpenMined).

4. Anonymity & Traffic-Analysis Resistance

Hide who is talking to whom, and the metadata around it.

• Tor — onion routing, the anonymity baseline.
• Nym — mixnet with cover traffic + incentives.
• I2P — garlic-routed anonymous network.
• Katzenpost — modern mix network.
• libsignal — the Signal Protocol (E2EE + metadata minimization).
• OpenMLS — Rust impl of IETF MLS (group E2EE).
• Oblivious HTTP (OHTTP, RFC 9458) — decouple requests from client identity.

5. Verified / Proof-Based Privacy

Prove something is true without revealing the underlying data.

• circom + snarkjs — zk-SNARK circuits + tooling.
• arkworks — Rust ecosystem for ZK / SNARKs.
• halo2 — recursive proof system (Zcash).
• gnark — Go zk-SNARK library.
• Noir — language for writing ZK circuits.

6. Hardware-Anchored Confidential Compute

Run code in an attestable, isolated enclave.

• Open Enclave SDK — cross-platform TEE SDK.
• Gramine — run unmodified apps in SGX.
• Occlum — memory-safe LibOS for SGX.
• Enarx — WebAssembly in TEEs (SGX/SEV).
• Constellation — confidential Kubernetes.

7. Accountability & Transparency

Make privacy claims auditable.

• Certificate Transparency / Trillian — verifiable, append-only logs.
• Sigstore — transparency for software signing.
• Warrant canaries — pattern + tooling (to be expanded).

---

Related Lists

Deep dives in a single discipline — this list maps the territory; these go deep:

• awesome-mpc · awesome-he · awesome-zero-knowledge-proofs · awesome-tor
• Consumer tools (different audience): awesome-privacy · Lissy93/awesome-privacy

Contributing

PRs welcome. Add a project with: name, link, and a one-line description of what privacy property it provides. Prefer maintained, documented, open-source projects. (CONTRIBUTING.md + license TODO.)

---

Curated by Cypherpunk School. Knowledge should be free.