This repository contains work on privacy-preserving smart contracts (PPSCs). We've included the 2 documents that cover zero-knowledge proof systems, Bulletproofs, and a rough outline on our vision for secure, private, flexible smart contracts. We wrote these in what we hope is an accessible way.
Previous talks on this subject were given at IEEE S&P 2020 (as a short talk; "The Marriage of Fully Homomorphic Encryption and Blockchain"), ZKSummit ("The Future of PPSCs"), and Devcon ("The Future of PPSCs").
We include a chart below for understanding what goes into our definitions of efficiency, security, privacy, and flexibility when discussing PPSCs. We call our project "Sunscreen" as our work will aim to satisfy 3 out of 4 of the goals defined—namely security, privacy, and flexibility—while sacrificing some degree of efficiency.
|Communication complexity?||Based on cryptography?||Confidential?||Easily adaptable to new environments(i.e. universal reference string)?|
|Reference string size?||Based on hardware?||Anonymous?||Supports arbitrary logic/computation?|
|Setup time?||Provable security?||Based on cryptography?||Supports "stateful" computation?|
|Time to generate transactions?||Non-standard assumptions?||Using tumblers or mixers?|
|Time to verify transactions?||Post-quantum?||Using stealth addresses?|
|Physical resources required?||Trusted setup?||Function privacy?|
|Potential for scalability?|
We provide a brief overview of the contents of the 2 documents below.
Disclaimer: These documents were originally written for internal usage and may no longer reflect up to date information or ideas.
We look at some efficient zero-knowledge proof protocols (i.e. SNARKs, STARKs, Bulletproofs) to explore which might be the most promising for a private transaction/smart contract scheme. We also evaluate some recent projects in the space including Zether, Hawk, Quisquis, Aztec, and Zexe.
We outline our design goals for a PPSC scheme and our core beliefs about what a "successful" scheme might look like for us. We provide a brief look into the building blocks we might use for such a scheme and the associated challenges around combining efficient ZKPs with FHE.
The appendix of this document was originally part of a separate internal document titled "Using Bulletproofs" from July 2019. In it, we consider two lines of work based on Bulletproofs. The first builds on Zether whereas the second argues for the creation of a new PPSC scheme using FHE inspired by some of the work in Short Discrete Log Proofs for FHE & Ring-LWE Ciphertexts.
This section includes protyping and benchmarking work for Short Discrete Log Proofs, and BGV. Current prototyping is being done in Julia by Bogdan for ease of iteration. However, we expect final libraries to be done in C/Rust.
A PoC of Short Discrete Log Proofs can be found here. We provide charts below on the performance of the verifiable encryption/decryption scheme of Section 1.5 of the paper. We stress that our prototype is not secure and should not be used in production.
Recall that the proofs use discrete logs and thus the performance is dependent on the curve chosen.
PoC; Curve25519; Intel i7 @ 2.6GHz
|Curve25519||1 thread||6 threads|
|Initial proof generation||2.15s||434ms|
PoC; secp256k1; Intel i7 @ 2.6GHz
|Secp256k1||1 thread||6 threads|
|Initial proof generation||16s||3.23s|
Encryption Scheme Performance
The encryption scheme is solely based on lattices (specifically ring-lwe) and thus not dependent on the curve choice.
|Ring-LWE Encryption Scheme||Timing (mean/median)|
BGV Scheme [WIP]
Benchmarking from HElib
FHE Requisite Knowledge
Even some of the nicest FHE libraries currently available are challenging to work with—requiring deep expertise in the scheme to achieve optimal performance. While both proponents and critics of FHE argue that FHE isn't being used due to efficiency reasons, we disagree as FHE can provide acceptable runtimes for certain use cases. Rather, we believe what's preventing FHE from gaining more traction is the level of expertise required to currently use the schemes/libraries.
Lack of Consistent Benchmarking
Additionally, there's a dearth of work in benchmarking some of the recent ZKPs and FHE schemes. It's difficult to incorporate (or even know if we should incorporate) these works when we don't know how they perform in practice. The efficiency tables in our PATS repo (such as this, and this) have many blanks as the authors have failed to provide important estimates in their papers. Many of the PPSC works also use vastly different machines—from an Intel i7-6820HQ throttled to 2.0 Ghz with <100MB of memory and a single thread to an Intel Xeon 6138 CPU at 3.0 Ghz with 252GB of RAM and 12 threads—further exacerbating the problem.
Some directions of research we'd like to pursue from here include:
- Tools for making FHE easier to work with
- Benchmarking ZKPs (possibly an entire project in and of itself!)
- Benchmarking FHE schemes (i.e. FV, BGV)
- Development of a user-friendly BGV library
Comments or Concerns
For any comments or concerns, feel free to open issues or ask questions on our Discord channel. Otherwise, you can contact Ravital (email@example.com) with general questions or for questions w.r.t. the research/documents. Questions about the specifics of the prototype/implementation can be directed to Bogdan (firstname.lastname@example.org).
Resources that are beginner friendly, we mark with
We suggest the following resources:
- ZKProof Community Reference, a fairly comprehensive document covering zero-knowledge proofs from the academic initiative ZKProof
- Bulletproofs notes, notes on how Bulletproofs work
- A Decade of Lattice Cryptography, a comprehensive document explaining the basics of lattices, assumptions used in lattice cryptography, and corresponding schemes
- Fundamentals of Fully Homomorphic Encryption - A Survey, an accessible but long introduction to FHE
- Computing Arbitrary Functions of Encrypted Data, a brief but well-written paper from Craig Gentry covering the basics of how FHE works
- A brief survey of Fully Homomorphic Encryption, a short and easy-to-read post covering the basic aspects of FHE and some of the challenges associated with it
- Microsoft's SEAL library, particularly the example sections to see how a FHE scheme works in practice