Skip to content

This lab enables Secure Chaincode Execution using Intel SGX for Hyperledger Fabric.

License

Notifications You must be signed in to change notification settings

mbrandenburger/fabric-private-chaincode

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Hyperledger Fabric Private Chaincode

Hyperledger Fabric Private Chaincode (FPC) enables the execution of chaincodes using Intel SGX for Hyperledger Fabric.

The transparency and resilience gained from blockchain protocols ensure the integrity of blockchain applications and yet contradicts the goal to keep application state confidential and to maintain privacy for its users.

To remedy this problem, this project uses Trusted Execution Environments (TEEs), in particular Intel Software Guard Extensions (SGX), to protect the privacy of chaincode data and computation from potentially untrusted peers.

Intel SGX is the most prominent TEE today and available with commodity CPUs. It establishes trusted execution contexts called enclaves on a CPU, which isolate data and programs from the host operating system in hardware and ensure that outputs are correct.

This project provides a framework to develop and execute Fabric chaincode within an enclave. It allows to write chaincode applications where the data is encrypted on the ledger and can only be accessed in clear by authorized parties. Furthermore, Fabric extensions for chaincode enclave registration and transaction verification are provided.

Fabric Private Chaicode is based on the work in the paper:

  • Marcus Brandenburger, Christian Cachin, Rüdiger Kapitza, Alessandro Sorniotti: Blockchain and Trusted Computing: Problems, Pitfalls, and a Solution for Hyperledger Fabric. https://arxiv.org/abs/1805.08541

This project was accepted via a Hyperledger Fabric RFC and is now under development. We provide an initial proof-of-concept implementation of the proposed architecture. Note that the code provided in this repository is still prototype code and not yet meant for production use!

For up to date information about our community meeting schedule, past presentations, and info on how to contact us please refer to our wiki page.

Architecture and components

Overview

This project extends a Fabric peer with the following components: A chaincode enclave that executes a particular chaincode, running inside SGX. In the untrusted part of the peer, an enclave registry maintains the identities of all chaincode enclaves and an enclave transaction validator that is responsible for validating transactions executed by a chaincode enclave before committing them to the ledger.

The following diagram shows the architecture:

Architecture

The system consists of the following components:

  1. Chaincode enclave: The chaincode enclave executes one particular chaincode, and thereby isolates it from the peer and from other chaincodes. A chaincode library acts as intermediary between the chaincode in the enclave and the peer. The chaincode enclave exposes the Fabric chaincode interface and extends it with additional support for state encryption, attestation, and secure blockchain state access. This code is executed inside an Intel SGX enclave.

  2. Enclave Endorsement validation: The enclave endorsement validation complements the peer’s validation system and is responsible for validating transactions produced by a chaincode enclave. In particular, the validator checks that a transaction contains a valid signature issued by a registered chaincode enclave. Iff the validation is successful, it causes the state-updates of the transaction to be committed to the ledger. This code is a normal Fabric transaction, i.e., executed and endorsed on multiple peers as required by the organization trust.

  3. FPC Chaincode Pkg: This component bundles together the chaincode enclave and the enclave endorsement validation logic into a fabric chaincode. It also includes a shim component which (a) proxies the chaincode enclave shim functionality, e.g., access to ledger, to the fabric peer, and (b) dispatches FPC flows to either the chaincode enclave (via __invoke queries) or to the enclave endorsement validation logic (via __endorse transactions).

  4. Enclave registry: The enclave registry (ercc) is a chaincode that runs outside SGX and maintains a list of all existing chaincode enclaves in the network. It performs attestation with the chaincode enclave and stores the attestation result on the blockchain. The attestation demonstrates that a specific chaincode executes in an actual enclave. This enables the peers and the clients to inspect the attestation of a chaincode enclave before invoking chaincode operations or committing state changes.

Design

More detailed architectural information and overview of the protocols can be found in the Fabric Private Chaincode RFC.

The full detailed operation of FPC is documented in a series of UML Sequence Diagrams. Note that FPC version 1.x corresponds to FPC Lite in documents and code.

Specifically:

  • The fpc-lifecycle-v2(puml) diagram describes the normal lifecycle of a chaincode in FPC, focusing in particular on those elements that change in FPC vs. regular Fabric.
  • The fpc-registration(puml) diagram describes how an FPC Chaincode Enclave is created on a Peer and registered in the FPC Registry, including the Remote Attestation process.
  • The fpc-key-dist(puml) diagram describes the process by which chaincode-unique cryptographic keys are created and distributed among enclaves running identical chaincodes. Note that in the current version of FPC, key generation is performed, but the key distribution protocol has not yet been implemented.
  • The fpc-cc-invocation(puml) diagram illustrates the invocation process at the beginning of the chaincode lifecycle in detail, focusing on the cryptographic operations between the Client and Peer leading up to submission of a transaction for Ordering.
  • The fpc-cc-execution(puml) diagram provides further detail of the execution phase of an FPC chaincode, focusing in particular on the getState and putState interactions with the Ledger.
  • The fpc-validation(puml) diagram describes the FPC-specific process of validation.
  • The fpc-components(puml) diagram shows the important data structures of FPC components and messages exchanged between components.
  • The detailed message definitions can be found as protobufs.
  • The interfaces document defines the interfaces exposed by the FPC components and their internal state.

Additional Google documents provide details on FPC 1.0:

  • The FPC for Health use case describes how FPC 1.0 enables a health care use case. The document also gives more details on the FPC 1.0-enabled application domains and related constraints. Lastly, it provides a security analysis why these constraints are sufficient for security.
  • The FPC externalized endorsement validation describes the FPC 1.0 enclave endorsement validation mechanism.

Source organization

  • client_sdk: The FPC Go Client SDK
  • cmake: CMake build rules shared across the project
  • common: Shared C/C++ code
  • config: SGX configuration
  • docs: Documentation and design documents
  • ecc_enclave: C/C++ code for chaincode enclave (including the trusted code running inside an enclave)
  • ecc: Go code for FPC chaincode package, including dispatcher and (high-level code for) enclave endorsement validation.
  • ecc_go: Go code for FPC Go Chaincode Support
  • ercc: Go code for Enclave Registry Chaincode
  • samples: FPC Samples
  • fabric: FPC wrapper for Fabric peer and utilities to start and stop a simple Fabric test network with FPC enabled, used by integration tests.
  • integration: FPC integration tests.
  • internal: Shared Go code
  • protos: Protobuf definitions
  • scripts: Scripts used in build process.
  • utils/docker: Docker images and their build process.
  • utils/fabric: Various Fabric helpers.

Releases

For all releases go to the Github Release Page.

WARNING: This project is in continous development and the main branch will not always be stable. Unless you want to actively contribute to the project itself, we advise you to use the latest release.

Getting started

The following steps guide you through the build phase and configuration, for deploying and running an example private chaincode.

We assume that you are familiar with Hyperledger Fabric; otherwise we recommend the Fabric documentation as your starting point. Moreover, we assume that you are familiar with the Intel SGX SDK.

This README is structure as follows. We start by cloning the FPC repository and explain how to prepare your development environment for FPC in Setup your FPC Development Environment. In Build Fabric Private Chaincode we guide you through the building process and elaborate on common issues. Finally, we give you a starting point for Developing with Fabric Private Chaincode by introducing the FPC Hello World Tutorial.

Clone Fabric Private Chaincode

Clone the code and make sure it is on your $GOPATH. (Important: we assume in this documentation and default configuration that your $GOPATH has a single root-directoy!) We use $FPC_PATH to refer to the Fabric Private Chaincode repository in your filesystem.

export FPC_PATH=$GOPATH/src/github.com/hyperledger/fabric-private-chaincode
git clone --recursive https://github.com/hyperledger/fabric-private-chaincode.git $FPC_PATH

Setup your Development Environment

There are two different ways to develop Fabric Private Chaincode. Using our preconfigured Docker container development environment or setting up your local system with all required software dependencies to build and develop chaincode locally.

Option 1: Using the Docker-based FPC Development Environment

In this section we explain how to set up a Docker-based development environment that allows you to develop and test FPC chaincode. The docker images come with all necessary software dependencies and allow you a quick start. We recommend to set privileges to manage docker as a non-root user. See the official docker documentation for more details.

First make sure your host has

  • Docker v23.0 (or higher). It also should use /var/run/docker.sock as socket to interact with the daemon (or you will have to override in $FPC_PATH/config.override.mk the default definition in make of DOCKER_DAEMON_SOCKET)
  • GNU make

Once you have cloned the repository, you can either use the pre-built images or you can manually build them. After that you will start the development container.

Pull docker images

To pull the docker image execute the following:

make -C $FPC_PATH/utils/docker pull pull-dev 

Manually build docker images

In order to build the development image manually you can use the following commands. Note that this process may take some time.

make -C $FPC_PATH/utils/docker build build-dev 

Start the dev container

Next we will open a shell inside the FPC development container, with environment variables like $FPC_PATH appropriately defined and all dependencies like the Intel SGX SDK, ready to build and run FPC. Continue with the following command:

make -C $FPC_PATH/utils/docker run-dev

Note that by default the dev container mounts your local cloned FPC project as a volume to /project/src/github.com/hyperledger/fabric-private-chaincode within the docker container. This allows you to edit the content of the repository using your favorite editor in your system and the changes inside the docker container. Additionally, you are also not loosing changes inside the container when you reboot or the container gets stopped for other reasons.

A few more notes:

  • We use Ubuntu 22.04 by default. To build also docker images with a different version of Ubuntu, add the following to $FPC_PATH/config.override.mk.
    DOCKER_BUILD_OPTS=--build-arg UBUNTU_VERSION=18.04 --build-arg UBUNTU_NAME=bionic
  • If you run behind a proxy, you will have to configure the proxy, e.g., for docker (~/.docker/config.json) and load the configuration inside the dev container by setting DOCKER_DEV_RUN_OPTS += -v "$HOME/.docker":"/root/.docker" in $FPC_PATH/config.override.mk. See Working from behind a proxy below for more information. Also note that with newer docker versions (i.e., docker desktop), the docker socket is located on the host in ~/.docker/. This may cause issues when using docker inside the FPC dev container as the docker client is not able to access the docker socket at the path of the host system. You may try to switch the docker context to use /var/run/docker.sock. We do not recommend this approach and happy for suggestions.
  • If your local host is SGX enabled, i.e., there is a device /dev/sgx/enclave or /dev/isgx and your PSW daemon listens to /var/run/aesmd, then the docker image will be sgx-enabled and your settings from ./config/ias will be used. You will have to manually set SGX_MODE=HW before building anything to use HW mode.
  • If you want additional apt packages to be automatically added to your container images, you can do so by modifying $FPC_PATH/config.override.mk file in the fabric-private-chaincode directory. In that file, define DOCKER_BASE_RT_IMAGE_APT_ADD_PKGS, DOCKER_BASE_DEV_IMAGE_APT_ADD_PKGS'and/or DOCKER_DEV_IMAGE_APT_ADD_PKGS with a list of packages you want to be added to you all images, all images where fabric/fpc is built from source and the dev(eloper) container, respectively. They will then be automatically added to the docker image.
  • Due to the way the peer's port for chaincode connection is managed, you will be able to run only a single FPC development container on a particular host.
  • For support for Apple Mac (M1 or newer) see the Troubleshooting section.

Now you are ready to start development within the container. Continue with building FPC as described in the Build Fabric Private Chaincode Section and then write your first Private Chaincode.

Option 2: Setting up your system to do local development

As an alternative to the Docker-based FPC development environment you can install and manage all necessary software dependencies which are required to compile and run FPC.

Requirements

Make sure that you have the following required dependencies installed:

  • Linux (OS) (we recommend Ubuntu 22.04, see list supported OS)

  • CMake v3.5.1 or higher

  • Go 1.21.x or higher

  • Docker 18.09 (or higher) and docker-compose 1.25.x (or higher) Note that version from Ubuntu 18.04 is not recent enough! To upgrade, install a recent version following the instructions from docker.com, e.g., for version 1.25.4 execute

    sudo curl -L "https://github.com/docker/compose/releases/download/1.25.4/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose	
    sudo chmod +x /usr/local/bin/docker-compose

    To install docker-componse 1.25.4 from docker.com, execute

    sudo curl -L "https://github.com/docker/compose/releases/download/1.25.4/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose
    sudo chmod +x /usr/local/bin/docker-compose
  • yq v4.x You can install yq via go get.

      go get github.com/mikefarah/yq/v4
  • Protocol Buffers

    • Protocol Buffers 3.0.x needed for the Intel SGX SDK
    • Protocol Buffers 3.11.x or higher and Nanopb 0.4.7
  • SGX PSW & SDK v2.22 for Linux (alternatively, you could also install it from the source

  • Credentials for Intel Attestation Service, read here (for hardware-mode SGX)

  • Intel Software Guard Extensions SSL (we recommend using tag 3.0_Rev2 OpenSSL 3.0.12)

  • Hyperledger Fabric v2.5.4

  • Clang-format 6.x or higher

  • jq

  • hex (for Ubuntu, found in package basez)

  • A recent version of PlantUML, including Graphviz, for building documentation. See Documentation for our recommendations on installing. The version available in common package repositories may be out of date.

Intel SGX SDK and SSL

Fabric Private Chaincode requires the Intel SGX SDK and SGX SSL to build the main components of our framework and to develop and build your first private chaincode.

Install the Intel SGX software stack for Linux by following the official documentation. Please make sure that you use the SDK version as denoted above in the list of requirements.

For SGX SSL, just follow the instructions on the corresponding github page. In case you are building for simulation mode only and do not have HW support, you might also want to make sure that simulation mode is set when building and installing it.

Once you have installed the SGX SDK and SSL for SGX SDK please double check that SGX_SDK and SGX_SSL variables are set correctly in your environment.

Protocol Buffers

We use nanopb, a lightweight implementation of Protocol Buffers, inside the enclaves to parse blocks of transactions. Install nanopb by following the instruction below. For this you need a working Google Protocol Buffers compiler with python bindings (e.g. via apt-get install protobuf-compiler python3-protobuf libprotobuf-dev). For more detailed information consult the official nanopb documentation http://github.com/nanopb/nanopb.

export NANOPB_PATH=/path-to/install/nanopb/
git clone https://github.com/nanopb/nanopb.git $NANOPB_PATH
cd $NANOPB_PATH
git checkout nanopb-0.4.7
cd generator/proto && make

Make sure that you set $NANOPB_PATH as it is needed to build Fabric Private Chaincode.

Moreover, in order to build Fabric protobufs we also require a newer Protobuf compiler than what is provided as standard Ubuntu package and is used to build the Intel SGX SDK. For this reason you will have to download and install another version and use it together with Nanopb. Do not install the new protobuf, though, such that it is not found in your standard PATH but instead define the PROTOC_CMD, either as environment variable or via config.override.mk to point to the new protoc binary

wget https://github.com/protocolbuffers/protobuf/releases/download/v22.3/protoc-22.3-linux-x86_64.zip
unzip protoc-22.3-linux-x86_64.zip -d /usr/local/proto3
export PROTOC_CMD=/usr/local/proto3/bin/protoc

Hyperledger Fabric

Our project fetches the latest supported Fabric binaries during the build process automatically. However, if you want to use your own Fabric binaries, please checkout Fabric 2.5.4 release using the following commands:

export FABRIC_PATH=$GOPATH/src/github.com/hyperledger/fabric
git clone https://github.com/hyperledger/fabric.git $FABRIC_PATH
cd $FABRIC_PATH; git checkout tags/v2.5.4

Note that Fabric Private Chaincode may not work with the Fabric main branch. Therefore, make sure you use the Fabric v2.5.4 tag. Make sure the source of Fabric is in your $GOPATH.

Build Fabric Private Chaincode

Once you have your development environment up and running (i.e., using our docker-based setup or install all dependencies on your machine) you can build FPC and start developing your own FPC application. Note by default we build FPC with SGX simulation mode. For SGX hardware-mode support please also read the Intel SGX Attestation Support Section below.

To build all required FPC components and run the integration tests run the following:

cd $FPC_PATH
make docker
make

Besides the default target, there are also following make targets:

  • build: build all FPC build artifacts
  • docker: build docker images
  • test: run unit and integration tests
  • clean: remove most build artifacts (but no docker images)
  • clobber: remove all build artifacts including built docker images
  • checks: do license and linting checks on source

Also note that the file config.mk contains various defaults which can all be redefined in an optional file config.override.mk.

See also below on how to build the documentation.

Intel SGX Attestation Support

To run Fabric Private Chaincode in hardware mode (secure mode), you need an SGX-enabled hardware as well corresponding OS support. However, even if you don't have SGX hardware available, you still can run FPC in simulation mode by setting SGX_MODE=SIM in your environment.

Note that the simulation mode is for developing purpose only and does not provide any security guarantees.

As mentioned before, by default the project builds in SGX simulation mode, SGX_MODE=SIM as defined in $FPC_PATH/config.mk and you can explicitly opt for building in hardware-mode SGX, SGX_MODE=HW. In order to set non-default values for install location, or for building in hardware-mode SGX, you can create the file $FPC_PATH/config.override.mk and override the default values by defining the corresponding environment variable.

Note that you can always come back here when you want a setup with SGX hardware-mode later after having tested with simulation mode.

Register with Intel Attestation Service (IAS)

If you run SGX in simulation mode only, you can skip this section. We currently support EPID-based attestation and use the Intel's Attestation Service to perform attestation with chaincode enclaves.

What you need:

  • a Service Provider ID (SPID)
  • the (primary) api-key associated with your SPID

In order to use Intel's Attestation Service (IAS), you need to register with Intel. On the IAS EPID registration page you can find more details on how to register and obtain your SPID plus corresponding api-key. We currently support both linkable and unlinkable signatures for the attestation.

Place your ias api key and your SPID in the ias folder as follows:

echo 'YOUR_API_KEY' > $FPC_PATH/config/ias/api_key.txt
echo 'YOUR_SPID_TYPE' > $FPC_PATH/config/ias/spid_type.txt
echo 'YOUR_SPID' > $FPC_PATH/config/ias/spid.txt

where YOUR_SPID_TYPE must be epid-linkable or epid-unlinkable, depending on the type of your subscription.

FPC Playground for non-SGX environments

FPC leverages Intel SGX as the Confidential Computing technology to guard Fabric chaincodes. Even though the Intel SGX SDK supports a simulation mode, where you can run applications in a simulated enclave, it still requires an x86-based platform to run and compile the enclave code. Another limitation comes from the fact that the Intel SGX SDK is only available for Linux and Windows.

To overcome these limitations and allow developers to toy around with the FPC API, we provide two ways to getting started with FPC.

  1. Using the Docker-based FPC Development Environment (works well on x86-based platforms on Linux and Mac).
  2. FPC builds without SGX SDK dependencies (targets x86/arm-based platforms on Linux and Mac).

We now elaborate on how to build the FPC components without the SGX SDK. Note that this is indented for developing purpose only and does not provide any protection at all.

In your config.override.mk set the following to variables:

FPC_CCENV_IMAGE=ubuntu:22.04
ERCC_GOTAGS=

This configuration sets a standard Ubuntu image as alternative to our fabric-private-chaincode-ccenv image and overrides the default build tags we use to build ercc.

Next you can build ercc using the following command:

GOOS=linux make -C $FPC_PATH/ercc build docker

For building a chaincode, for instance $FPC_PATH/samples/chaincode/kv-test-go, just run:

GOOS=linux make -C $FPC_PATH/samples/chaincode/kv-test-go with_go docker

You can test your FPC chaincode easily with one of the sample deployments tutorials. We recommend to start with the-simple-testing-network.

Notes:

  • On Mac use a recent version of bash (brew install bash).
  • TODO more to come

Troubleshooting

This section elaborate on common issues with building Fabric Private Chaincode.

Docker

Building the project requires docker. We do not recommend to run sudo make to resolve issues with mis-configured docker environments as this also changes your $GOPATH. Please see hints on docker installation above.

The makefiles do not ensure that docker files are always rebuild to match the latest version of the code in the repo. If you suspect you have an issue with outdated docker images, you can run make clobber build which forces a rebuild. It also ensures that all other download, build or test artifacts are scrubbed from your repo and might help overcoming other problems. Be advised that that the rebuild can take a fair amount of time.

Working from behind a proxy

The current code should work behind a proxy assuming

  • you have defined the corresponding environment variables (i.e., http_proxy, https_proxy and, potentially, no_proxy) properly, and
  • docker (daemon & client) is properly set up for proxies as outlined in the Docker documentation for clients and the daemon.
  • the docker version is correct. Otherwise you may run into problems with DNS resolution inside the container.
  • the docker-compose version is correct. For example, the docker-compose from Ubuntu 18.04 (docker-compose 1.17) is not recent enough to understand ~/.docker/config.json and related proxy options.

Furthermore, for docker-compose networks to work properly with proxies, the noProxy variable in your ~/.docker/config.json should at least contain 127.0.0.1,127.0.1.1,localhost,.org1.example.com,.example.com.

Another problem you might encounter when running the integration tests insofar that some '0.0.0.0' in integration/config/core.yaml used by clients -- e.g., the peer CLI using the address: 0.0.0.0:7051 config as part of the peer section -- result in the client being unable to find the server. The likely error you will see is err: rpc error: code = Unavailable desc = transport is closing. In that case, you will have to replace the '0.0.0.0' with a concrete ip address such as '127.0.0.1'.

Environment settings

Our build system requires a few variables to be set in your environment. Missing variables may cause make to fail. Below you find a summary of all variables which you should carefully check and add to your environment.

# Path to your SGX SDK and SGX SSL
export SGX_SDK=/opt/intel/sgxsdk
export SGX_SSL=/opt/intel/sgxssl

# Path to nanopb
export NANOPB_PATH=$HOME/nanopb

# SGX simulation mode
export SGX_MODE=SIM

# SGX simulation mode
export SGX_MODE=HW

The file config.mk contains various defaults for some of these, but all can be (re)defined also in an optional file config.override.mk.

Clang-format

Some users may experience problems with clang-format. In particular, the error command not found: clang-format appears even after installing it via apt-get install clang-format. See here for how to fix this.

ERCC setup failures

If, e.g., running the integration tests executed when you run make, you get errors of following form:

Error: endorsement failure during invoke. response: status:500 message:"Setup failed: Can not register enclave at ercc: Error while retrieving attestation report: IAS returned error: Code 401 Access Denied"

In case you run in SGX HW mode, check that your files in config/ias are set properly as explained in Section Intel Attestation Service (IAS). Note that if you run initially in simulation mode and these files do not exist, the build will create dummy files. In case you switch later to HW mode without configuring these files correctly for HW mode, this will result in above error.

no Raft leader

The following error message sometimes appears when running the integration tests in the $FPC_PATH/integration folder. The output contains the following:

got unexpected status: SERVICE_UNAVAILABLE -- no Raft leader

Rerunning the tests usually works. If this error appers during the make step of building FPC than uncommenting some integration tests fixes the issue.

Working with the FPC dev container

To make starting and stopping the dev container more reliable it is advised to use the following commands:

  • Start the container and get a shell: make -C $FPC_PATH/utils/docker run-dev
  • Get another shell inside the dev container: docker exec -it fpc-development-main /bin/bash
  • Stop the container: docker stop fpc-development-main

Development on Apple Mac (M1 or newer)

For developers using Apple Mac (M1 or newer) we suggest to use the prebuilt FPC dev container. Add the following configuration to your config.override.mk, pull the docker images and start the FPC dev container as described above in Option 1: Using the Docker-based FPC Development Environment. Note that SGX is not supported on Apple platforms, and hence, FPC chaincode can only be used in simulation mode. Alternatively, a cloud-based development environment can be used with SGX HW support, see our tutorial How to use FPC with Azure Confidential Computing.

DOCKER_BUILD_CMD=buildx build
DOCKER_BUILD_OPTS=--platform linux/amd64
DOCKER_DEV_RUN_OPTS=--platform linux/amd64

Building Documentation

To build documentation (e.g., images from the PlantUML .puml files), you will have to install java and download plantuml.jar. Either put plantuml.jar into in your CLASSPATH environment variable or override PLANTUML_JAR or PLANTUML_CMD in config.override.mk (see config.mk for default definition of the two variables). Additionally, you will need the dot program from the graphviz package (e.g., via apt-get install graphviz on Ubuntu).

By running the following command you can generate the documentation.

cd docs
make

Developing with Fabric Private Chaincode

In the samples folder you find a few examples how to develop applications using FPC and run them on a Fabric network. In particular, samples/application contains examples of the FPC Client SDK for Go. In samples/chaincode we give illustrate the use of the FPC Chaincode API; and in samples/deployment we show how to deploy and run FPC chaincode on the Fabric-samples test network and with K8s (minikube).

More details about FPC APIs in the Reference Guides Section.

Your first private chaincode

Create, build and test your first private chaincode with the Hello World Tutorial.

Developing and deploying on Azure Confidential Computing

We provide a brief FPC on Azure Tutorial with the required steps to set up a confidential computing instance on Azure to develop and test FPC with SGX hardware mode enabled.

Reference Guides

Management API

While the management API for Fabric is mostly unchanged, some modifications are needed for FPC to work. In particular, FPC extends the Fabric's lifecycle API with additional commands to create an FPC enclave and handle the key provisioning. These are detailed separately in the FPC Management API document

FPC Shim

The FPC Shim follows the programming model used in the standard Fabric Go shim and offers a C++ based FPC Shim to FPC chaincode developers. It currently comprises only a subset of the standard Fabric Shim and is complemented in the future. These details are documented separately in the Shim header file itself: ecc_enclave/enclave/shim.h

Important: The initial version of FPC, FPC 1.0 (aka FPC Lite), has a few constraints in applicability and programming model. Hence, study carefully the section discussing this in the FPC RFC and the comments at the top of shim.h before designing, implementing and deploying an FPC-based solution.

FPC Client SDK

In order to interact with a FPC chaincode you can use the FPC Client SDK for Go or use the Peer CLI tool provided with FPC. Both make FPC related client-side encryption and decryption transparent to the user, i.e., client-side programming is mostly standard Fabric and agnostic to FPC.

The FPC Client SDK for Go is located in client_sdk/go. See also Godocs.

For the command-line invocations, use the $FPC_PATH/fabric/bin/peer.sh wrapper script. We refer to our integration tests for usage examples.

Getting Help

Found a bug? Need help to fix an issue? You have a great idea for a new feature? Talk to us! You can reach us on Discord in #fabric-private-chaincode.

We also have a weekly meeting every Tuesday at 3 pm GMT on Zoom. Please see the Hyperledger community calendar for details.

Contributions Welcome

For more information on how to contribute to Fabric Private Chaincode please see our contribution section.

References

  • Marcus Brandenburger, Christian Cachin, Rüdiger Kapitza, Alessandro Sorniotti: Blockchain and Trusted Computing: Problems, Pitfalls, and a Solution for Hyperledger Fabric. https://arxiv.org/abs/1805.08541

  • Fabric Private Chaincode RFC

  • Presentation at the Hyperledger Fabric contributor meeting August 21, 2019. Motivation, background and the inital architecture. Slides

  • Presentation of at the Hyperledger Fabric contributor meeting November 11, 2020. The design and rationale for FPC Lite (FPC 1.0). Slides

Project Status

Hyperledger Fabric Private Chaincode was accepted via a Hyperledger Fabric RFC and is now under development. Before, the project operated as a Hyperledger Labs project. This code is provided solely to demonstrate basic Fabric Private Chaincode mechanisms and to facilitate collaboration to refine the project architecture and define minimum viable product requirements. The code provided in this repository is prototype code and not intended for production use.

Initial Committers

Core Team FPC 1.0

Sponsor

Gari Singh ([email protected])

License

Hyperledger Fabric Private Chaincode source code files are made available under the Apache License, Version 2.0 (Apache-2.0), located in the LICENSE file.

About

This lab enables Secure Chaincode Execution using Intel SGX for Hyperledger Fabric.

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages

  • Go 42.4%
  • C 16.3%
  • C++ 15.8%
  • Shell 14.6%
  • Makefile 4.7%
  • CMake 3.7%
  • Dockerfile 2.5%