In agricultural society for thousands of years, the determinants of economic development are land and labor.In the industrial age, factors such as capital and technology have gradually become more important social resources.When the era of the digital economy comes, cutting-edge technologies such as big data artificial intelligence continue to emerge, and global data is showing an explosive growth trend. The value of massive data in the fields of AI, finance, and medical care has become a consensus among all parties. Data has just become this era.The core production factor drives human society to a higher stage of development.
However, unlike production factors such as land, capital, and technology, data will leak specific information once it is “seen”, making it difficult to restrict usage and usage, making data from all parties reluctant to share, data fragmentation, data barriers andThe data island phenomenon is serious.At the same time, the fact that data is easily copied indefinitely makes the value difficult to measure, and cannot be priced or traded through market supply and demand, which hinders the large-scale and orderly circulation of data as a factor of production in the market.
Specifically in real life, data has been regarded as an “invisible asset” by all walks of life. Many companies are actively exploring how to solve the contradiction between data circulation and privacy data protection through technical means, so that they can makeData can truly become a kind of “invisible asset” for added value.At present, the solution found in the industry is to use privacy calculation to help all parties only obtain the value of the data when using the data, avoid exposing the original data, and make the data “available and invisible”.According to 36 Krypton, since 2018, the number of companies providing privacy computing solutions has gradually increased, and the attention of the capital market has also increased with market trends.
Huakong Qingjiao Information Technology (Beijing) Co., Ltd. (hereinafter referred to as Huakong Qingjiao) is one of the well-known companies in the industry.The company is an information technology company launched by Tsinghua University in June 2018, focusing on research, development and construction of modern cryptography and game theory-based big data security fusion technology, standards and infrastructure.Former Goldman Sachs global partner Mr. Zhang Xudong became CEO, and Professor Xu Wei of Tsinghua University served as chief scientist.
At present, the company has nearly 100 employees, of which more than 75% have doctorate and master degrees. The employees come from Huawei, Ali, Baidu, IBM, Google, Oracle and Thomson Reuters.The company’s main shareholders include Tsinghua University, China Internet Finance Association, Beijing Haidian District Entrepreneurship Support Fund, Hong Kong Exchanges and Clearing Limited (Hong Kong Stock Exchange), Lenovo Group and Gao Rong Capital.
On the product side, the company has independently developed and launched a multi-party secure computing (MPC)-based data security fusion platform, which can enable multiple non-trusted databases to perform efficient data fusion calculations on the premise of confidential data.The platform uses strict mathematically proven cryptography theory to replace ciphertext calculations with ciphertext calculations at the computer instruction set and compiler level. The calculation contract between the data provider and the data consumer is used to allocate computing power to execute the privacy calculation service, andThe calculation result is given to the result acquirer specified in the contract, thereby truly realizing “the data is available and invisible” and “specifying the purpose and usage of the data” (“calculation contract”).
In addition, the data fusion platform developed by Huakong Qingjiao is a universal platform with enterprise-level deployment, convenient development, carrying technology and computing scalability.At the same time can also provide a flexible combination of multi-level products and services.
Huakong Qingjiao’s data fusion platform
In fact, as early as the 1980s, Turing Award winner Academician Yao Qizhi used mathematical theory to prove that all calculations that can be performed on plain data can theoretically be calculated directly on the cipher text and obtained and calculated in plain text.The completely consistent results have created a multi-party safe computing theory.
However, there is often an insurmountable gap between theory and reality.In a live broadcast sharing a few days ago, Zhang Xudong, the CEO of Huakong Qingjiao, said that although the security of multi-party secure computing has been verified by mathematical theory very early, compared with plaintext computing, the calculation of full ciphertext has a greater impact on computing power and performance.The requirements are more than a million times higher.How to achieve multi-party secure computing cryptography theory to privacy computing technology is a huge challenge worldwide.
According to him, Huakong Qingjiao has gone through a series of engineering research explorations and practices, based on the realization of multi-party secure computing technology of data “available and invisible”, combined with federal learning, trusted computing, data desensitization and differential privacyEtc. Based on plaintext data privacy protection technology and blockchain certificate storage technology, a set of privacy computing technology system based on the combination of ciphertext computing and plaintext computing based on multi-party secure computing was developed and created, which realized the practicality of multi-party secure computing technology andProductization has improved the performance of ciphertext computing to a level that can meet commercial applications.
Zhang Xudong believes that ciphertext calculation and plaintext calculation are not a competitive relationship, but a complementary relationship.When solving the problem of computational efficiency, it is necessary to combine plaintext computation and ciphertext computation. The former guarantees efficiency, and the latter guarantees data privacy and security at the most critical places.The flexible use of the two can balance the accuracy, confidentiality and efficiency of the calculation to the greatest extent.
From cryptography theory to general privacy protection computing technology system
In terms of “regulated usage and usage”, Zhang Xudong further explained that the prescribed usage is to specify the algorithm, and the prescribed usage is to limit the number or frequency of use.In the process of data calculation, the data user and the data provider reach a usage agreement according to the demand before initiating the calculation task, and the calculation is performed according to the agreed agreement. This is also referred to as the “calculation contract” mentioned above.When the computing task is completed under the constraints of the contract, the data provider shares the specific right to use the data without exposing any clear text information.Huakong Qingjiao said that by establishing a cryptography theory to a general privacy protection computing system, it can provide a technical grasp for the elementalization of data and lay the foundation for the construction of data interconnection, integration and circulation infrastructure.
According to 36 Krypton, there have been many researches and explorations on privacy computing technology in the industry, and each has its own advantages and disadvantages.There are also many companies in the industry that provide solutions related to privacy computing. Due to the differences in background and technology accumulation, each company’s reliance on various technical routes is also different.The practical requirements of the safe integration of data in China must be a comprehensive solution that combines multiple technologies.
Currently commonly used privacy computing technology
36 Krypton found through an interview that the current mainstream technical route of privacy computing can be divided into multi-party secure computing (MPC), trusted execution environment (TEE), and federal learning.Huakong Qingjiao currently provides a comprehensive privacy computing solution, which is widely used in various scenarios such as query, joint statistics, and joint modeling.Other related companies in the industry also include Shuyu Technology, Yifang Jianshu, Chongwei Technology, etc. BAT also has its own solutions.
In the live broadcast, Zhang Xudong, CEO of Sinotrans Qingjiao analyzed the understanding of the factorization of data production based on industry trends, focusing on the reasons, difficulties, how to do and the prospect of the future of “data becomes a factor of production”.The redefinition of the “wave-particle duality” of data value proposes how to use the calculated value of data to solve the development bottleneck of traditional information sharing by adopting a new method of data security fusion in data circulation.36 krypton has also been organized according to related content.
Data becomes the premise of production factors
Zhang Xudong believes that with the development level of productivity, different historical periods have different core production factors, the agricultural society is land and labor, the industrial society is capital, technology and management, the information age is knowledge, and today’s digital economy era, data is informationThe objects and result carriers of processing and calculation are the basis for decision-making to optimize the allocation of natural and social resources. The use of data for overall monitoring, regulation and local optimization of social and natural resources can greatly improve social labor.productivity.This is that data “natural” is the core element of the digital economy.
However, data will not become “natural” as a factor of production, and it will become a factor of production only after solving two core problems.One is circulation, and the other is confirmation of power.
Data has different characteristics from other production factors: low copy cost, can be copied indefinitely, can be used by multiple parties at the same time, and new data can be produced during the simultaneous use.Traditional information is based on the sharing of plaintext data, and the plaintext data will leak specific information once it is seen. It is difficult to limit its use and usage, and it is difficult to clarify “responsibility, power, and profit.”Large-scale market circulation.
In response to these, Zhang Xudong proposed the “wave-particle duality” of data, that is, the value of data lies in its visible information value on the one hand, similar to the “particles” of quantum mechanics; on the other hand, the value of the result of its calculation, Ie its computational value, is similar to the “wave” of quantum mechanics.
The meaning of the data “available and invisible” and the prescribed usage of the calculated value
Now that big data and artificial intelligence are widely used, the value of data is more reflected in its computing value.Separating the information value and calculation value carried by the data to make the data “available and invisible” can avoid the unlimited supply and use of plain text data caused by being seen.
At the same time, it is necessary to specify the specific use and usage of the calculated value of the data. Only in this way can the qualitative and quantitative supply of data be realized.Therefore, it is not the data itself that sets prices through the market and circulates on a large scale, but the specific right to use the data.The data “available and invisible” and “prescribed use and usage” can realize the separation of data ownership, use right, beneficiary right and disposal right, and lay the foundation for confirming the right of the data to become the production factor.
Zhang Xudong also emphasized the concept of “data fusion”, that is, when multiple data are used together, the internal dimensions of the data can be superimposed, which greatly increases (1+1>2) the computational value of the data.The calculated value of the data is reflected through the calculation. Multi-party data is calculated according to the prescribed use through the “available and invisible” method. The calculation result produced by each algorithm is a new value, and the data for each calculation isOne consumption of a specific right of use.”Data fusion” can effectively solve the previous problems of data “unable to share”, “unwilling to share”, and “dare not to share”.
Prospect of data realization of production factors
Looking to the future, Zhang Xudong used the “data call” to make an analogy to the blueprint for data to realize the production factor.Using “telephone” as an analogy to client data service (DS), DS will be spread across every social individual that provides or uses specific usage rights for data, and “program-controlled exchange” as a metaphor for privacy computing services. In “telephone” and “program-controlled exchange”What flows between them is the calculation factor-the ciphertext fragment of the data, which itself does not carry any “visible” information.”Program-controlled switches” and “program-controlled switches” will also be connected into a multi-dimensional national data network, which can connect every social individual.
Basic modules of data elementization infrastructure
On the whole, Zhang Xudong believes that contract-based privacy computing technology can effectively build a closed loop of socialized data, and truly eliminate the concerns of data attribution, data security, and privacy protection in different parts of the data value chain.Mass production lays the technical foundation for data to truly become a factor of production.