Recently, at the 2022 Privacy Computing Conference, the China Academy of Information and Communications Technology announced the results of the sixth batch of trusted privacy computing evaluations. The Meituan Privacy Computing Platform passed the two special evaluation certifications of "Federal Learning Security" and "Multi-Party Secure Computing Basic Capabilities". In 2021, Meituan has passed the special evaluation and certification of "Federal Learning Basic Ability".
Passing these three certifications marks that the data security circulation technology solution provided by the Meituan Privacy Computing Platform (the product name is "Sifang Street") can serve financial, medical, government affairs, marketing and other scenarios through multi-party secure computing and federated learning capabilities. According to the data fusion requirements under the following conditions, it supports data demanders, data suppliers, and algorithm parties to complete data application, authorization, analysis, modeling, and prediction online to ensure that data is available and invisible.
What is certification?
Certification is to prove that the product meets the standard. Therefore, product certification needs to have standards first. The test standards for the two certifications passed this time are "Security Requirements and Test Methods for Privacy Computing Federated Learning Products" and "Technical Requirements and Test Methods for Data Circulation Products Based on Multi-Party Secure Computing".
Certification mark : certification will issue a certification certificate, or allow the product to use the certification mark, and some certification marks will be affixed on the product or its packaging to explain to the society that it has passed the standard test and meets the requirements of the standard and product performance.
"Federal Learning Security Special Project", the detection content includes general algorithm protocol security, security intersection protocol security, feature engineering protocol security, joint modeling protocol security, joint prediction algorithm protocol security, AI attack resistance, password security, communication security, authorization Security, system security, stability, storage security, logs and certificates, a total of 13 special evaluations, comprehensively inspect the security of the federated learning system.
The basic capabilities of multi-party secure computing . The detection content includes evaluation items including privacy set intersection, hidden information retrieval, joint statistics, and general secure multi-party computing compilation. It strictly meets the evaluation requirements in terms of product security, robustness, and stability.
What is private computing?
At the National Two Sessions in 2022, "Digital Economic Governance" first appeared in the "Government Work Report", and privacy computing became a new focus. Privacy computing is a series of information technologies that realize data value mining on the premise of ensuring that the data provider does not disclose the original data. It provides a "available and invisible" solution for the circulation of data value.
From a principle point of view, privacy computing is an interdisciplinary technical system that integrates cryptography, data science, artificial intelligence, blockchain and many other fields, including three technical solutions: multi-party secure computing MPC, federated learning FL, Letter Execution Environment TEE.
From the perspective of application, privacy computing is gradually advancing from the financial field to the development of many fields such as medical care and government affairs. Privacy computing first has a large area of landing and application in finance, including bank risk control, credit business and anti-fraud, etc. At the same time, it has also started in various scenarios such as medical auxiliary diagnosis, government data opening, smart city and Internet of Things. Gradually explore and land.
The realization of "available and invisible" mainly relies on privacy computing
The privacy computing core solves the problems of data sharing, data credibility, and data ownership. In the current digital economy, the importance of data elements has been gradually recognized, but how to carry out data circulation and governance still faces two major problems:
-
Compliance flow . With the gradual improvement of privacy-related regulations, data can no longer be transmitted and circulated in clear text as before. Enterprises need to realize data flow under the premise of protecting privacy and meeting compliance requirements.
-
Data Ownership . Enterprises regard data as their most important asset and are unwilling to share data with other organizations. Or in some enterprises, because of data security and privacy, data between different departments are not willing to open to each other.
These problems can all be solved through privacy computing. The core capability of privacy computing enables all parties to realize the circulation of data value without the original data leaving the domain. On the one hand, it solves the compliance problem of data integration, on the other hand, it protects the data ownership of the data owner, and realizes the "usable but invisible" data.
In the future, Meituan will continue to deepen the exploration of the application of the combination of privacy computing technology and Meituan's business, make full use of the advantages of the technology platform, activate the value of data fusion in a safe and compliant way, and help Meituan's various businesses to quickly carry out extensive and comprehensive cooperation with industry partners , jointly guarantee the security of users' personal information, and promote the efficient and benign flow of data elements.
---------- END ----------
read more
Frontend | Algorithms | Backend | Data
Security | Android | iOS | Operation and Maintenance | Testing