Skip to content

WeDPR v3.0.0

Pre-release
Pre-release
Compare
Choose a tag to compare
@cyjseagull cyjseagull released this 16 Dec 02:57
· 1 commit to main since this release
v3.0.0
f41d97a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.

获取更多信息，请阅读WeDPR 3.x文档

新增

WeDPR隐私计算平台站点端管理功能，包括：

丰富的隐私计算任务支持

  • 多方(>=2)隐私求交集
  • 匿踪查询任务
  • 多方(>=2)联合建模任务(SecureBoost, SecureLR)
  • 多方(>2)数据联合分析任务(隐私SQL, 隐私Python)

基于区块链灵活可扩展的多方同步模块

  • 基于区块链进行跨机构元数据信息同步

丰富的数据源管理

  • 支持HDFS, Hive, MYSQL和国产数据库等多种数据源接入

细粒度的用户和权限体系

  • 支持多用户模式，并支持用户维度的数据、服务权限管理

通用的数据、服务审批流

  • 实现了审批模块，支持数据、服务授权

基于Jupyter和wedpr-ml-toolkit sdk的专家模式

在向导模式的基础上，为数据和模型开发人员提供了基于Jupyter的专家模式，便于其获取建模、隐私求交集的结果，完成后续数据分析工作。

  • 实现了用户维度的Jupyter管理，多用户的Jupyter环境通过linux用户体系完全隔离开
  • 提供了wedpr-ml-toolkit工具，便于用户在Jupyter专家模式环境中发起隐私求交集、联合建模等隐私计算任务，并可灵活地获取任务执行结果

支持DAG的任务调度模块

  • 任务调度模块支持DAG工作流
  • 各类隐私计算节点多活可扩展

服务发布功能

  • 数据集可发布为匿踪查询服务，供授权的用户或机构查询
  • 联合建模训练过程中产生的模型可发布为服务，供联合预测使用

API接入

  • 支持应用方通过申请的凭证AccessKey接入到管理平台

What's New

WeDPR privacy computing platform site management, features include:

Rich support for privacy computing tasks

  • Multi-party (>=2) privacy intersection
  • Privacy query task
  • Multi-party (>=2) joint modeling task (SecureBoost, SecureLR)
  • Multi-party (>2) data joint analysis task (privacy SQL, privacy Python)

Flexible and scalable multi-party synchronization module based on blockchain

  • Cross-agency metadata information synchronization based on blockchain

Rich data source management

  • Supports multiple data source access such as HDFS, Hive, MYSQL and domestic databases

Fine-grained user and permission system

  • Supports multi-user mode and user-level data and service permission management

General data and service approval process

  • Implemented the approval module to support data and service authorization

Expert mode based on Jupyter and wedpr-ml-toolkit SDK

Based on the wizard mode, an expert mode based on Jupyter is provided for data and model developers to obtain modeling and privacy intersection results and complete subsequent data analysis work.

  • Implemented user-level Jupyter management, and the multi-user Jupyter environment is completely isolated through the Linux user system
  • The wedpr-ml-toolkit tool is provided to facilitate users to initiate privacy computing tasks such as privacy intersection and joint modeling in the Jupyter expert mode environment, and can flexibly obtain task execution results

Support DAG task scheduling module

  • Task scheduling module supports DAG workflow
  • Various privacy computing nodes are multi-active and scalable

Service publishing

  • Datasets can be published as privacy query services for authorized users or institutions to query
  • Models generated during the joint modeling training process can be published as services for joint prediction

API Access

  • Supports application parties to access the management platform through the applied credential AccessKey
Assets 4
Loading