qiyuangong / confidential-computing-zoo

Confidential Computing Zoo provides confidential computing solutions based on Intel SGX, TDX, HEXL, etc. technologies.

Home Page:https://cczoo.readthedocs.io/en/latest/index.html

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool


Confidential Computing Zoo (CCZoo) is a collection of code-ready reference solutions, which can be used as a copy-paste developer guide, demonstrating how to apply modern security technologies to real-life cloud business scenarios, in order to facilitate the developers to build their own end-to-end Confidential Computing solutions more easily. Some of the solutions are also validated on the public cloud services, such as Alibaba Cloud, Tencent Cloud, AWS, Azure, etc. Please see Cloud Deployment.

The concerned modern security technologies are (but not limited to): TEE (Trusted Execution Environment, such as Intel® SGX and TDX), HE (Homomorphic Encryption) and its hardware accelerations, Remote Attestation, LibOS, cryptographic and its hardware accelerations. The concerned business scenarios are (but not limited to): cloud native AI inference, vertical and horizontal federated learning, big data analytics, key management, RPC (Remote Process Call, such as gRPC), etc.

CCZoo maintains a live table, as below, to indicate the correlations between business usages (rows) and security technologies (columns). Each hyperlink will direct you to the document section that explains the corresponding details and then guides you to the source codes. Enjoy!

Solution List (Solution to Component Correlation)

  Solution                                                     Security Components

Validated
in Public Cloud

Status

  TEE

           LibOS

Remote Attestation

    KMS

HE

Crypto

TLS

SGX TDX Gramine Occlum *RATS-TLS *RA-TLS gRPC Vault eHSM-KMS
Multi-Party Compute / Federated Learning
Horizontal Federated Learning
(
TensorFlow)
Yes - Yes - - Yes
(2-way)
- - - Yes Yes
(RA-gRPC)

Alibaba Cloud,
Tencent Cloud,
ByteDance Cloud

Published
Vertical Federated
Learning
 
(
TensorFlow)
Yes - Yes - - Yes
(2-way)
- - - Yes Yes
(RA-gRPC)

Alibaba Cloud,
Tencent Cloud,
ByteDance Cloud

Published
Private Set
Intersection 
Yes - Yes - - - - - - - - - In Progress
Secure Logistic
Regression Training
Base on TEE & 
HE 
Yes - Yes - - - - - Yes Yes Yes Alibaba Cloud,
Tencent Cloud
Waiting For Publish
Secure AI Inference & Training
TensorFlow Serving
Cluster PPML

(TensorFlow, K8S)
Yes - Yes Yes - - - - - Yes Yes

Alibaba Cloud,
Tencent Cloud,
ByteDance Cloud

Published
Leveled HE Logical Regression Inference - - - - - - - - Yes - - - In Progress
Secure BigDL
Recommend System
- Yes - - - - - - - - - - Not Start
Native Application Hosting
Cross Language
framework Based
on Gramine
Yes - Yes - - - - - - - - Tencent Cloud Published
Attestation Server & Key Management Service
Attestation and Secret Provision Service Yes Yes - - Yes Yes - Yes - Yes Yes - Published
eHSM-KMS Yes - - - - - - Yes - Yes Yes - Published
Optimization on Secure Libs
Private Set
intersection
Optimization
on Xeon
- - - - - - - - Yes Yes - - Not Start
Secure Database
Secure Database
Querying Based
on HE
- - - - - - - - Yes Yes - - Not Start

Incubating Component Projects

Besides reference solutions, CCZoo is also incubating new projects of key security components that are commonly used by multiple CCZoo reference solutions. Once any of them is proven useful enough and stable enough via a thorough validation with CCZoo reference solutions running on various public cloud services, it will graduate from CCZoo and evolve to a standalone project.

Incubating Component Project '*'                                                                         Description Status Validated in Public Cloud
RATS-TLS This project provides a proof-of-concept implementation on how to integrate Intel SGX and TDX remote attestation into the TLS connection setup. Conceptually, it extends the standard X.509 certificate with SGX and TDX related information. It also provides two non-SGX clients (Wolfssl and OpenSSL) to show how seamless remote attestation works with different TLS libraries.  Published Alibaba Cloud
RA-TLS Enhanced gRPC This project provides an enhanced gRPC (Remote Procedure Call) framework to guarantee security during transmission and runtime via two-way RA-TLS (Intel SGX Remote Attestation with Transport Layer Security) based on TEE (Trusted Execution Environment). Published Alibaba Cloud,
Tencent Cloud,
ByteDance Cloud

Cloud Deployment

Solutions and incubating component projects in CCZoo are constantly extended to be validated in public clouds to verify the versatility, stability, robustness. We will provide detialed configurations of each public clouds for reference, and notes of the diversity in each cloud for easy delopyment.

Below table shows solutions and component projects validated in public clouds. And it will be updated continuously.

     Public Cloud Alibaba Cloud ByteDance Cloud Tencent Cloud
Instance  Type g7t
ecs.ebmg2t.32xlarge
M6ce.4XLARGE128 
Kernel 4.19.91-24
kernel-5.15
5.4.119-19-0009.1
OS Alibaba Cloud Linux 2.1903 Ubuntu20.04 TencentOS Server 3.1
Memory 64G(32G EPC memory)
512GB(256GB EPC memory)
64G(32G EPC memory)
vCPU 16 16 16
PCCS Server sgx-dcap-server.cn-hangzhou.aliyuncs.com sgx-dcap-server-tc.sh.tencent.cn 
Validated Solution 

Confidential Computing Zoo Documentation

The official confidential computing zoo documentation can be found at https://cczoo.readthedocs.io.


Community Involvement

  • Please submit issues in this project if there is any question or request.
  • Welcome PRs for contributions.

Welcome to join the Wechat group or Slack channel for CCZoo tech discussion.

You can check CCZoo previous PDT meeting munites here.

About

Confidential Computing Zoo provides confidential computing solutions based on Intel SGX, TDX, HEXL, etc. technologies.

https://cczoo.readthedocs.io/en/latest/index.html

License:Apache License 2.0


Languages

Language:CMake 45.0%Language:C++ 17.4%Language:Python 15.1%Language:C 8.4%Language:Shell 6.0%Language:Cython 3.6%Language:Dockerfile 2.5%Language:Makefile 1.2%Language:Java 0.5%Language:Batchfile 0.1%Language:HCL 0.0%