Text | Wei Qiyang
Source | Intelligent Relativity (aixdlun)
When data becomes an important factor of production, how should data privacy protection and data use be weighed?
In April of this year, the "Opinions of the Central Committee of the Communist Party of China and the State Council on Building a More Complete Factor Market Allocation System and Mechanism" (hereinafter referred to as the "Opinions") was released. The data as a new type of production Traditional factors such as labor force, capital and technology are listed as one of the factors.
The "Opinions" clarified that we should accelerate the cultivation of the data element market, promote the open sharing of government data, enhance the value of social data resources, and strengthen data resource integration and security protection.
The value of data is being re-recognized, so how to guard the security of data?
The era of data elements is coming, and privacy computing usher in an explosion node
In fact, as early as October-November last year, the National Development and Reform Commission awarded six regions including Xiong'an, Zhejiang, Fujian, Guangdong, Chongqing, and Sichuan as "Digital Economy Innovation and Development Pilot Zones" to explore the digital economy and various industries. Fusion development.
Earlier, Internet applications represented by WeChat, short videos, and live broadcasts changed our lives; e-commerce platforms represented by Taobao, JD.com, and Pinduoduo changed our consumption patterns; and Li Jiaqi and Wei Ya as examples The representative cargo anchor has innovated new business formats; the "new infrastructure" power represented by 5G and industrial Internet is driving the transformation of productivity...
Behind all this is the support of data elements.
Statista data shows that in 2020, the global big data market revenue is expected to reach 56 billion U.S. dollars, an increase of about 33.33% from the expected level in 2018, and double the market revenue in 2016. The global data volume reached 41ZB in 2019 and is estimated to reach 50.5ZB in 2020.
Global big data market data volume and market revenue scale (data source: China Academy of Information and Communications Technology, compiled by CB Insights China)
On the other side of data value, data security and privacy protection have become contradictions.
For example, many government data cannot be released to the public. Customer data collected by communication operators, banks, Internet companies, etc. is subject to legal regulations and cannot be disclosed to third parties. Therefore, data is divided into islands and data cannot be communicated. , The value of data is difficult to reflect.
However, even so, we receive various marketing calls, or sell real estate, or recommend stocks almost every day. In the face of various black properties, there is no personal privacy at all. Every year, data and privacy leaks exposed by various industries are endless. The public has a great distrust of the process of data value generation.
The ten largest data breaches in the world in 2019 (arranged by CB Insights China)
Due to its reproducibility and dissemination, data cannot be shared safely in nature. However, in the era of data elements, although we do not share the data itself, the value of the data should be shared.
To solve this contradiction, privacy computing technology has emerged.
Privacy computing, according to the definition of China Academy of Information and Communications Technology, refers to the information technology that analyzes and calculates data and can verify the results of the calculation on the premise that the data provider does not leak sensitive data.
In a broad sense, it refers to computing systems and technologies oriented to privacy protection, covering the entire information flow process of data generation, storage, calculation, application, and destruction. The desired effect is to make data "available and invisible" in all links.
To put it in a more general way, it is to allow data to be freely circulated or shared under the premise of ensuring data security, eliminating the problem of data islands, thereby releasing greater value of data, improving production efficiency, and promoting industrial innovation.
From this point of view, how big the big data market is, how high the ceiling of privacy computing is.
Born for the value of data, the four little dragons of privacy computing emerge
Privacy computing is a comprehensive technology, specifically, it mainly includes three directions.
One is the multi-party secure computing (MPC) technology based on cryptography. Through special encryption algorithms and protocols such as secret sharing, forgetting transmission, obfuscating circuits or homomorphic encryption, it supports direct calculation on encrypted data. Theoretically, under the "ideal" situation that does not consider the cost, the multi-party secure computing technology can achieve arbitrary computing "functions" and achieve relatively high security. However, due to the sudden increase in data communication volume, large loss of computing efficiency and the need for extremely high computing power requirements, the technical productization of MPC still has certain restrictions, and related technical solutions are actively exploring.
The second is the federated learning technology based on artificial intelligence. In the horizontal dimension, each participant trains and calculates his own sample locally, and only shares the gradient of model training; in the vertical dimension, each participant trains their own embedding ("vector mapping") to jointly train the upper model. The integration of the two dimensions allows multiple data owners who do not trust each other to jointly conduct model training without sharing data.
The third is the secure sandbox computing (TEE) technology based on trusted hardware. The core idea is to build a hardware security area, data is only calculated in the security area, use the trusted execution environment TEE to prevent the operating system from maliciously viewing the content of the application execution environment; use the security sandbox to prevent malicious applications from controlling operations through special calls system.
At present, the general consensus in the industry is that in order to realize data "available but not visible", it is difficult for a single technology to take the lead. The complementary integration of different technology paths (cryptography, artificial intelligence, blockchain, etc.) is the development trend.
It is also in this context that the four privacy computing players of Ant Financial, WeBank, Huakong Clearing, and Yifang Jianshu ran out of a group of competitors, relying on their comprehensive strengths. The solution became the "four little dragons" of the privacy computing track.
For example, Ant Golden uses the TED ENGINE engine for data security and privacy protection, which integrates three technologies: sensitive data intelligent marking technology (Tag), AI security enhancement technology (Enhace) and intelligent threat recognition technology (Detection). In Enhace technology, Ant Financial focuses on differential privacy and trusted hardware.
In addition, Ant Financial has also developed a large-scale multi-party secure computing commercial platform-Morse, which directly provides personalized multi-party secure computing services to other enterprises and organizations to solve practical business problems.
Based on the integration of cryptographic algorithms, privacy protection algorithms, secure multi-party computing and other technologies, WeBank has developed a set of instantaneous privacy protection solutions WeDPR. At this year's Hangzhou Blockchain International Week, WeDPR was rated as "the most powerful application in privacy computing scenarios".
It is also worth mentioning that Fate, the alliance learning open source project of WeBank, also has independent intellectual property rights for federal learning.
Huakong Qingjiao specializes in multi-party security computing. The founder Yao Qizhi is the dean of the Institute of Interdisciplinary Information of Tsinghua University and the only Chinese winner of the Turing Award. The technical theory of multi-party secure computing originated from the "Millionaire" vision proposed by Yao Qizhi in the 1980s. As the proponent and important founder of this theory, he is committed to promoting the implementation of technology. Its PrivPy platform implements a high-performance universal secure computing framework, clustering and scalable solutions.
Yifangjianshu's data privacy computing platform Yishufang takes "Internet of Data and Computing" (IoDC) as its core to build an open ecosystem, and has landed the country's first large-scale private computing platform in Xiamen to realize its data strategy The "landing" from top-level design to bottom-level implementation. The platform not only integrates self-developed technology, but also integrates excellent third-party technical solutions, including homomorphic encryption, blockchain, federated learning, etc., through computing to achieve data connection, sharing and value realization.
On another level, from the perspective of the industry of privacy computing applications, finance and medical are the two most important tracks.
We understand the importance of data security to the financial industry very well. When it is small, it is related to each of us’s pockets. When it is large, it is related to the country’s economic foundation. Therefore, we can see that privacy computing is the "four little dragons". "In, Ant Financial, WeBank, and Huakong Clearance are all on the financial track.
Data security is equally important to the medical industry. Jin Tao, associate professor of the School of Software at Tsinghua University, said in an interview with the media at the 2019 Big Data Industry Summit that health and medical data not only involve the personal level, but also the public interest, and even national security. For example, if a person suffers from an epidemic or infectious disease, his personal data may involve the optimization and improvement of the entire treatment plan, which will benefit the whole society; genetic data may be related to national security.
Yifang Jianshu has taken a different route from the other three, focusing its business on the medical track that also has a greater demand for privacy computing.
In this way, it is easy to understand that not only the superior technology is used as the supporting base, but also the preemptive position in the main track is realized. The above constitutes the moat of the privacy computing "four little dragons" in the market competition.
To bring comprehensive data value, privacy computing still needs to do three things
Although the current privacy computing industry presents a competitive landscape of "four superpowers and more powerful", the value of data has not been fully discovered, and privacy computing technology is far from reaching its end. In the future era of data elements, look for comprehensive data value , Privacy computing still needs to do three things.
1. Business landing: More industry applications to ensure the "robustness" of data generalization applications
Privacy computing is currently mainly implemented in three scenarios: finance, medical treatment, and marketing. In the future, it will certainly penetrate into more industries and scenarios. This requires technology to be very robust. After changing the environment, the system /Technology is also the same ability as in the previous environment.
For example, an open privacy computing platform can meet the needs of a financial institution well, and after switching to another financial institution, it can also quickly adjust to meet the individual needs of the institution; the scope is expanded a bit, the privacy When the computing platform is switched to the medical industry, it can also run well and has comprehensive capabilities.
In fact, although the privacy computing "Four Little Dragons" are focused on the financial and medical industries, they continue to penetrate into other industries. With medical care as an entry point, Yifang Jianshu has a layout in finance, marketing, insurance, government affairs and other industry scenarios. The greatest significance of cross-industry pan-data applications is to break data islands and maximize the value of data.
2. Achievement transformation: wider coverage, data application from a single enterprise to the entire city
At present, data circulation is basically "self-produced and self-sold" within a single enterprise. Data security is guaranteed by building a data platform, but the value of data is difficult to output. When the data island is eliminated, the value of data should circulate in an unlimited space. That is, the leap from a single enterprise to all industries and entire cities.
City-level landing, Yifang Jianshu has a case that can be used for reference. They built a medical big data application and open platform based on private computing technology in Xiamen, which is the first known case of using private computing technology to achieve city-level applications. .
More importantly, because the same organization can play multiple roles on the platform, such as government agencies, medical companies, etc., which can provide a large amount of raw data, and there is a demand for medical data, the value of data has been reshaped. To a certain extent, Yifang Jianshu's scientific research transformation in Xiamen has also become a turning point for the value of medical data circulation.
3. Strategic height: a more open ecology, to achieve an upgrade and leap in play thinking
The business referred to by privacy computing now is more of a solution for enterprises. The value of data is single. Everyone knows how to circulate data, but there is no clear direction on how to circulate it.
Here, the ecological play often mentioned in the Internet industry may become a way of thinking. The circulation of anything needs to be in a mature ecosystem, and all parties involved in the ecology can benefit from it in order to make circulation more efficient. Data circulation certainly does not deviate from this rule.
The "Internet of Data and Computing (IoDC)" currently being established by Yifang Jianshu mentioned above is quite a rudimentary form of a privacy computing ecology in terms of its gameplay thinking. There are three ecosystems to be established under it: data interconnection An ecosystem where ecology, data and artificial intelligence algorithms are interconnected, an ecosystem where data providers, data users and data service providers coexist.
According to the information disclosed by Luo Zhen, the founder and CEO of Yifang Jianshu, in his previous speech, Yifang Jianshu has already joined forces in vertical fields such as WeBank and Huakong Clearing to participate in the IoDC network in terms of data ecology. Construct.
In this way, some companies have begun to think and explore the data ecology in depth, instead of pulling the value of the data in their hands, but in an ecologically open way, so that every ecological participant can get benefits from it.
The more the merrier.
The open ecosystem is more effective in terms of technology development and market development, and will become the mainstream of the privacy computing industry in the foreseeable future.
Dig the smart well, and add vx: zhinengxiaoyan
This content is the original [Intelligent Theory of Relativity],
It only represents personal views. Without authorization, no one may use it in any way, including reprinting, extracting, copying, or creating mirror images.
Some pictures are from the Internet, and the copyright ownership has not been verified. They are not for commercial use. If there is any infringement, please contact us.
Intelligent Relativity (WeChat ID: aixdlun):
• AI industry new media;
• Today’s Toutiao Qingyun Project winners TOP10;
• The paper's monthly top5 in the news technology list;
• Articles "occupies" the TOP10 list of popular articles in titanium media for a long time;
• Author of "One Hundred Thousand Whys of Artificial Intelligence"
• [Key areas of focus] Smart home appliances (including white appliances, black appliances, smart phones, drones and other AIoT devices), smart driving, AI+ medical care, robotics, Internet of Things, AI+ finance, AI+ education, AR/VR, cloud computing , Developers and the chips and algorithms behind them.