Multiple technology paths have evolved together, and MPC, tee and federal learning are at the present stage. Privacy computing has many underlying technologies to choose from, and the implementation of privacy computing may require the integration of multiple technologies. At this stage, MPC, tee and federal learning technologies are leading the commercialization process, and this technology trend will continue in the short term. In the long run, MPC and federated learning require privacy computing providers to accumulate effective data and iterate and optimize algorithms for a long time, while tee needs to optimize the design of the underlying chip on this basis. In general, tee has extremely high requirements for the full stack capacity of software and hardware of suppliers. At this stage, Chinese manufacturers can only realize it by Internet head manufacturers. In consideration of cost, the application proportion of MPC and federal learning may increase.
It is difficult for enterprises to establish a privacy computing ecosystem with a single competitive advantage, and both algorithms and data are indispensable. Privacy computing solutions are highly dependent on the maturity of their underlying technologies. However, only with the underlying technology of privacy computing, enterprises can not provide effective services for users. The needs of users are often very customized and specialized for the needs of data. Therefore, privacy computing providers need to provide users with effective desensitization data. On the other hand, without a large amount of effective data, the evolution speed and optimization degree of privacy computing algorithm will also be affected. Therefore, the establishment of privacy computing ecosystem requires suppliers to master excellent algorithms and rich data resources. Internet manufacturers, start-ups deep into vertical scenes and leading enterprises in the industry have these two resources, which makes it possible for them to successfully layout private computing.
The generality of privacy computing is limited to the underlying technology, and the vertical scene algorithm has high discrimination. The underlying technology of privacy computing is universal, and the establishment of the underlying platform is possible. Looking at the development ecology of privacy computing, the main players can be divided into three categories: Internet enterprises, enterprises in vertical industries and start-ups. Internet companies are most likely to develop universal underlying platforms. The future ecology may further develop vertical solutions based on different scenarios based on this underlying platform.
The degree of open source of privacy computing technology is limited, and manufacturers with priority layout have the first mover advantage. At this stage, China’s privacy computing service providers only open source Tencent, Weizhong bank, Baidu, byte beat and matrix element. Tencent’s open source framework is gradually adopted at the bottom of the Internet, and Tencent’s open source framework is gradually adopted. Generally speaking, the openness of the source code of private computing is low, and the manufacturers who give priority to layout have advantages on the algorithm side. Late entrants can also realize the rapid upgrading of algorithms with their own data and ecological resources. The latecomers with richer ecological resources and more traffic entrances have a relatively higher probability of success. There is another possibility for the future development trend, that is, when the openness of private computing is enhanced or the head supplier establishes the underlying platform of private computing, some start-ups with human resources will develop special and vertical solutions based on the underlying platform and occupy a certain market share.
The underlying technology of privacy computing is universal, and the establishment of the underlying platform is possible. Since 2018, leading Internet enterprises, China United Network Communications Limited(600050) , China Telecom Corporation Limited(601728) , China Mobile and other communication operators, mature network security and big data companies such as Fushu, tongdun and xinghuan, as well as start-up technology enterprises such as Huakong qingjiao, Weiwei and insight have successively joined the Bureau of privacy computing. It is recommended to pay attention to the operators of layout privacy computing technology, China Mobile (600941, not rated), China United Network Communications Limited(600050) (600050, not rated), China Telecom Corporation Limited(601728) (601728, not rated).
Risk tips
The research and development progress of secure multi-party computing, federal learning, confidential computing and other technologies is less than expected; Intensified market competition, investment suggestions and investment targets
