Point stone privacy calculation

Based on the technical capabilities of multi-party secure computing, federated learning, trusted execution environment (TEE), blockchain, etc., under the premise of protecting data security and privacy protection, it meets the requirements of multi-party data security sharing, opening, integration, modeling and computing, enables customers in government affairs, finance, medical care, education, automobile, Internet and other fields, and solves the dilemma of data fusion applications.

Apply for privatization version Apply for SaaS version

Product Overview
Product Matrix
Product advantages
Product Functions
Application scenarios
Customer Stories
Related products

Product Overview

It supports multi-party data security "integration" and modeling, analysis and calculation, provides full process services from data analysis, model training, evaluation to application deployment, and has different cloud/end service deployment modes. Solve the business needs of multiple scenarios, such as risk control, precision marketing and cold start of new users, break the isolated island of data applications, effectively guarantee data security and user privacy, and achieve data availability, invisibility, security compliance, flow and sharing.

Product Matrix

Baidu Dianshi Union Learning

 The distributed architecture, based on MPC, FL, DP and other technologies, provides data security and privacy protection capabilities for the whole process of data joint analysis, joint computing and joint modeling.

Baidu Dianshi Confidential Calculation

 A general secure computing platform based on trusted hardware provides high-performance data flow services and privacy computing services.

Baidu Dianshi Data Security Sandbox

 A secure and controllable data security sharing platform to assist the government and institutions in solving data security issues and privacy protection issues in the process of data opening.

Product advantages

Technological advantages of independent research and development

The independently developed heterogeneous privacy computing solution fully adapts to the information innovation system; It has more than 300 domestic and foreign patents and more than 20 top conference papers; Participated in the formulation of industry standards related to privacy computing for many times.

Rich open forms

Member of the Privacy Computing Alliance of the ICT Academy, and member of the CCC Confidential Computing Alliance of the Linux Foundation; Initiator of PaddleFL and Apache Teachave privacy computing and confidential computing open source projects.

Mature industry experience

Mature delivery experience in government affairs, biomedical, transportation, manufacturing and other fields; The rich model training experience covers old customer mining, access, anti fraud, black industry, risk control and other scenarios.

Product Functions

Security data fusion
Multiple security guarantees
Provide data security fusion solutions based on federated learning, secure multi-party computing, data security sandbox, trusted computing and other technologies.
Self developed technology
The independently developed heterogeneous privacy computing solution ensures the security of data storage and computing while making data available and invisible.
Multiple modeling methods
Form based interaction
Users can easily complete model construction and iteration through form based interaction.
Personalized modeling
The product also supports personalized modeling analysis and tuning in a secure computing environment.
Rapid service deployment
Rapid deployment model
It supports the rapid deployment and release of the trained model, and can be seamlessly connected to different business application systems of customers.
Service performance monitoring
Provide service performance monitoring to ensure service stability; It also supports dynamic capacity expansion.

Application scenarios

Financial anti fraud

Marketing insight

Open government data

Scenario Overview

The current financial environment is complex, and the risk of financial fraud is rising. However, the traditional anti fraud means are single, inefficient, and limited in scope, making it difficult to cope with the emerging new fraud means.

We can provide

Leading technology

Support multi-party data security fusion and modeling calculation on the basis of ensuring data security

Whole process service

Full process services from model training, assessment to application deployment to help financial customers identify fraud risks in the whole business process

Related products

Number privacy service PNS

Customer Stories

Big data engineering service project of a provincial capital city

Computing nodes are deployed locally in nearly 50 commission bureaus and social institutions to meet the demands of all parties for "data not leaving the domain", "available and invisible", and network connectivity, provide government data infrastructure support for government brain and business applications, and realize the safe sharing and opening of government data.

Joint modeling and precision marketing of an airline company

Set up the Dianshi federal learning platform, safely integrate the data model training of both parties, predict the flight upgrade high net worth people, and push it to the customer marketing platform, reach people through information flow, SMS and other channels to achieve joint precision marketing. Totally 300w model calls have been completed, and the customer conversion rate has increased significantly.

Big data engineering service project of a provincial capital city

A government&enterprise data platform project

By building a pointstone federal learning platform, joint computing/joint modeling can be carried out without data from all parties leaving the domain. Standardize the gathering, sharing and open development process of government and enterprise data resources, and achieve orderly, open and open government data resources.

State Key Laboratory of Media Communication Content Cognition

By building a pointstone data security sandbox platform, it provides TensorFlow, Python, BML and other machine learning frameworks. On the premise of ensuring data security, it reduces the threshold of machine learning, maximizes the value of data development, and realizes the value of production, learning, research and use.