[go: nahoru, domu]

Skip to content

A system for verifying outsourced computations, and applying SNARKs. Simplified release of the main Pepper codebase.

License

Notifications You must be signed in to change notification settings

jasonge27/pequin

 
 

Repository files navigation

Pequin: An end-to-end toolchain for verifiable computation, SNARKs, and probabilistic proofs

Pequin is a toolchain to verifiably execute programs expressed in (a large subset of) the C programming language. There are two parties: a prover and verifier. Using Pequin, the prover can convince the verifier that it executed a given computation according to the program expressed by the verifier. More generally, using Pequin, the prover can convince the verifier of the truth of some assertion, without the verifier having to manually check every line of the proof of that assertion.

Pequin consists of a front-end and a back-end. The front-end takes C programs and transforms them to a set of arithmetic constraints, in such a way that the constraints are satisfiable if and only if the purported output (which will be produced by the prover, and which is represented by variables in the constraints) is what the original program would have produced. The back-end is a probabilistic proof protocol by which the prover convinces the verifier that the constraints are indeed satisfiable (if they are); if the constraints are not satisfiable, the verifier is not fooled, except with vanishingly low probability.

The specific back-end in Pequin is a zk-SNARK (zero-knowledge succinct non-interactive argument of knowledge). Pequin uses SCIPR Lab's libsnark, which is an optimized implementation of the back-end of Pinocchio, itself a refinement and implementation of GGPR.

Pequin is a result of several years of research in the Pepper project, done at NYU and UT Austin. Pepper has brought to bear powerful techniques from complexity theory and cryptography: probabilistically checkable proofs (PCPs), efficient arguments, interactive proofs (IPs) etc. Pepper's research results include reducing the computational costs of a PCP-based efficient argument by over 20 orders of magnitude (in base 10!), and extending verifiability to representative uses of cloud computing (MapReduce jobs, simple database queries, computations involving private databases, etc.). These results and others are published in peer-reviewed scientific papers.

Pepper itself has code; the goal of Pequin in particular is to be easier to use, and more composable. Development is ongoing. Please open issues or submit pull requests! (This source code is released under a BSD-style license. See LICENSE for more details.)

What's coming next?

  • Incorporating other back-ends
  • Improving documentation and usability of this codebase is ongoing.

Installation and first steps

  1. This codebase depends on several external libraries. Installation scripts are provided for a few common linux distros. For example, on Ubuntu 14.04 or Debian Jessie,

    ./install_debian_ubuntu.sh
    

    will install some packages from apt repositories and compile our snapshots of some third-party dependencies. (You will be prompted to enter your password to install the apt packages as sudo.)

  2. ./install_buffet.sh will build the patched Clang/LLVM libraries for Buffet's C-to-C compiler, needed for running computations with data-dependent loops and control flow.

For more information on setting up dependencies (for example, for other distros), see INSTALLING.md and compiler/buffetfsm/README.md.

Once everything is set up, you're ready to run some verifiable computations! Please see GETTINGSTARTED.md for a quick overview of the process.

Contact

Please contact pepper@pepper-project.org for any questions and comments. We are happy to work with you to adapt this technology into your application.

About

A system for verifying outsourced computations, and applying SNARKs. Simplified release of the main Pepper codebase.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C 43.1%
  • Go 42.2%
  • Java 9.2%
  • C++ 3.7%
  • Python 1.3%
  • Shell 0.2%
  • Other 0.3%