In July 2022, China officially announced the 2022 privacy computing technology innovation figures selected.
Among the selected candidates this time, we have seen research scholars who uphold scientific research pragmatism, applied figures who have achieved technological innovation, and industry figures who have achieved industrial innovation. They are seeking new opportunities and constantly promoting a new look in technological research and industry applications. We will send exclusive interviews from selected candidates one after another, enter their innovative achievements and share their understanding and experience of privacy computing.
privacy computing has developed rapidly in recent years. IDCh data shows that "in 2021, China's privacy computing market size has exceeded 860 million yuan, and it is expected to achieve a market growth rate of more than 110% in the future." It is a multi-disciplinary technology that involves cryptography , system security, and machine learning. Specific scenario applications also involve sub-fields such as medical care and finance. Therefore, this field requires cross-domain "all-rounders".
Wang Shuang has been deeply engaged in the field of privacy computing for more than ten years. Thanks to his high interdisciplinary background in applied physics, biomedical engineering , and electronics and computer engineering, he has not only understood privacy computing, but also understood his application scenarios in the medical field, and has also promoted his repeated technological breakthroughs and innovations.
In 2011, he pioneered the federated learning framework and applied it to medical online learning, combining privacy computing with medical care, and completing the development of underlying frameworks and algorithms. This framework serves multiple national-level medical and health networks and is also a breakthrough at the framework level of the federated learning system. This technology has helped to connect the data interconnection of five hospitals under the system of the University of California, realize "data available and invisible" across multiple hospitals, and help multi-center joint analysis and research. After
, based on the concept of federated learning, combined with different technologies such as homomorphic encryption, multi-party secure computing, and trusted execution environment, Wang Shuang led the team to achieve larger-scale implementation applications, realizing multi-center joint computing research and transnational rare disease research across multiple institutions under the premise of privacy protection, as well as cross-border rare diseases research across multiple institutions under the premise of 30 million people.
at that time supported multi-center genomic data analysis in different countries such as the United States, the United Kingdom, Singapore , as well as different analyses such as horizontal and vertical federal learning, structured data, unstructured data, genetic data, imaging data, etc.
As of now, Wang Shuang has published more than 100 international academic works and has received tens of millions of US dollars in the Natural Science Foundation. Deeply pursuing the academic field may be one of Wang Shuang’s choices in his career path, but he did not stop there.
In 2014, Wang Shuang and other experts founded the iDASH Privacy Protection Computing Competition (including tracks such as homomorphic encryption, multi-party secure computing and trusted execution environment). As of now, the competition has been held to the ninth session. It has attracted high attention from universities, startups and large manufacturers around the world to privacy computing. At the same time, the competition also promoted the cultivation of cross-field talents.
Wang Shuang said, "What is exciting is that can see the significant improvement in efficiency or performance of this technology in the iDASH Privacy Protection Computing Competition every year. As the development of privacy computing becomes more mature, it can slowly support more practical application scenarios."
Figure丨"MIT Technology Review" China's 2022 Privacy Computing Technology Innovation Figure Wang Shuang
2018 is of great significance to Wang Shuang. At that time, he was a professor in the Department of Biomedical Information at the University of California, San Diego School of Medicine and was selected as the "Overseas High-Level Young Talent" of the Organization Department of the National Central Committee.
With the development of privacy computing, Wang Shuang realized that in addition to focusing on technological innovation and following up on academic research, the influence of privacy computing in society should also be expanded."Only by implementing and popularizing large-scale applications in business scenarios can we better promote the influence of privacy computing on a larger level," said Wang Shuang.
coincides with the rapid development of China's digital economy. Wang Shuang decided to return to China to establish Tsingwei Technology, so that the commercial development of privacy computing can "become and blossom" in the market of the motherland. Under his leadership, the team has implemented cross-multi-country rare disease association analysis for the first time in the world that simultaneously supports federated learning, trusted execution environment, homomorphic encryption and multi-party secure computing, and supports converged solutions for different privacy protection technologies.
Currently, Wang Shuang serves as the Distinguished Professor of West China Hospital of Sichuan University and the Visiting Professor of Affiliated Hospital of Tongji University. At the same time, he is also the founder, chairman and CTO of Derwei Technology.
10 billion market driven by policies, laws and markets
In recent years, privacy computing has developed comprehensively. This is driven by the multi-faceted driving of policies, laws and markets. According to iResearch Consulting data, "By 2025, China's privacy computing market size will reach 14.51 billion yuan."
From the policy level, 2020, data was written into national policies as a new production factor. It is worth noting that the right to use, ownership and management of data is difficult to separate. If data is used as a factor of production and the data source continues to create value, a technical means is needed to enable data to realize value on time, quantity and purpose.
privacy computing, as a technical solution, can fully reflect the factorized value of data through the "invisible, controllable and measurable" method.
From a legal perspective, Internationally, EU began to implement the "HTML General Data Protection Regulation" (GDPR for short) in 2018, and the United States also promulgated the "California Consumer Privacy Act of 2018, referred to as CCPA). China followed closely behind, promulgating the " Data Security Law " and the " Personal Information Protection Law " at the end of 2021. These legal provisions clearly stipulate civil liability, criminal liability and related compensation liability corresponding to improper data protection or privacy leakage. Among them, the liability for compensation stipulated in the GDPR and the Personal Information Protection Law is 4% and 5% of annual income respectively. Therefore, the many compliance risks at the legal level have also given rise to a large demand for privacy computing in data interaction scenarios.
From the market level, is currently in the era of big data and AI. The development of AI technology is no longer just a competition of models, but a matter of who can reach more data, including data volume and data dimension information. Therefore, to a large extent, the reachability of data determines the capabilities of AI algorithms.
With the surge in demand for big data, the data volume and data dimensions of each single center can no longer be met due to its limitations, which in turn has triggered the demand for multi-center cooperation. Under the multi-faceted driving and demand of policies, laws and markets, privacy computing can meet the joint analysis of multi-center compliance, and also conform to the trend of technological development.
3 Deeply cultivated "privacy computing + medical care" and completed multiple benchmark cases
In recent years, Derwei Technology has gradually become a typical representative of "privacy computing + medical care" enterprises. "Providing a safer solution through secure federated learning is not only limited to federated learning, but also we also integrate technologies such as trusted execution environment, multi-party secure computing, homomorphic encryption, etc. to further protect the gradient information security of federated learning interaction and the security of model results." said Wang Shuang .
So, what are the "secret weapons" of the technological advancement of Darwei Technology? On the one hand, the team has been focusing on the underlying technology research of privacy computing for more than ten years and has a profound industry knowledge-How. On the other hand, the difference between Derwei Technology and most privacy computing companies is that it has done relatively deep medical scenarios.
Wang Shuang said, "We have a multi-technology integration system that can automatically orchestrate the underlying privacy computing platform and framework of different technical routes. This helps us better integrate different technologies and solve the needs of different performance, security levels, accuracy and other needs in real scenarios."
So, what is the difference between medical scenarios and non-medical scenarios? In fact, in addition to the structured data covered by non-medical scenarios, it also includes its unique unstructured data, genomic data, medical imaging data, etc. The main difference between the two is the data type and the required data processing methods. Obviously, the types that need to be processed in medical scenarios are more complex and require hundreds of different analysis methods.
Wang Shuang pointed out: "In medical scenarios, the requirements for multi-center concurrency are relatively high. In most non-medical scenarios, data source cooperation is usually two to three parties. When doing scientific research or new drug development in medical scenarios, it may require the joint participation of more than a dozen or even hundreds of hospitals. Therefore, this also requires that the underlying platform needs to support large-scale and hundreds of concurrent joint computing."
From the perspective of accuracy, medical scenarios involve the life safety of patients. In order to avoid the difficult problem of medical responsibility accidents, privacy calculations in this scenario also require that does not introduce additional calculation errors, so a lossless privacy calculation solution is required.
pic丨Wang Shuang (Source: Wang Shuang)
In recent years, Wang Shuang has led the team to complete several benchmark cases. For example, and several medical institutions have completed the world's first cross-multi-country rare disease data analysis with privacy protection, supports converged solutions with different privacy protection technologies, and supports multi-center secure federated learning of horizontal and vertical multimodal medical data.
Wang Shuang said, "Research on rare diseases of multinational multi-center data collaboration systems with privacy protection is often limited by insufficient data volume in a single center, and the cross-center flow of biomedical data is still subject to legal supervision and high requirements for security measures."
To solve this problem, Wang Shuang and his team rely on the Privacy Protection Computing Platform, developed a multinational multi-center data collaboration system with privacy protection for it to analyze the genetic data of Kawasaki disease in children. uses this system to perform secure distributed computing on encrypted data, solving the problem of difficulty in cross-border flow of medical data and ensuring that all data sharing complies with the regulatory requirements of data flow in various countries.
It is understood that during this process, personal privacy data and intermediate results will not be disclosed, whether intentionally or unintentionally. "It is worth noting that the platform will not introduce significant computing load restrictions, which greatly improves the feasibility of secure large-scale cross-border genetic data analysis in practice," said Wang Shuang.
In addition, The team also conducted the first cross-provincial multi-center whole genome analysis in the country, and developed a real-time monitoring and early warning system for new and sudden infectious diseases based on multi-dimensional big data. enables data to be safely shared and interconnected, and helps with precision medicine, drug research and development, tiered diagnosis and treatment , epidemic prevention and control and other fields.
makes data "move" to form a network effect, and builds privacy computing into a technology and security base in the digital economy era
Regarding the recent development of 阿日本 Technology, Wang Shuang said that on the one hand, 阿日本 Technology empowers applications in different industries with "privacy computing +" and provides an underlying and core privacy computing platform. Continuous investment and innovation will be made in technology, including core technicians, financial investment and external cooperation. "We will pay great attention to the combination of production, education, research and application, and continuously improve the security of the entire system, the performance, accuracy, flexibility and the overall competitiveness of the system."
On the other hand, the team will continue to invest and strengthen cooperation in the construction of data resources. Through technological leadership, more nodes have the ability to have privacy computing, and then each node forms a data network.
In addition, in order to help data sources convert data value, Wang Shuang and his team will build more open applications and ecology with data users or data application provider , to "use" the data source and let the data "move" to form a network effect.
pic丨Wang Shuang (Source: Wang Shuang)
In Wang Shuang's view, the privacy computing platform can serve as a technology and security base in the digital economy era. Its essential value lies in comprehensively releasing the value of data, helping users reduce costs and increase efficiency, improving data usage efficiency, and enhancing mutual trust in the process of data interaction. "Privacy computing may serve as an infrastructure in the future to serve various scenarios that require data interaction, similar to the mobile Internet that people now support with 4G and 5G base stations."
He said that privacy computing will become an underlying infrastructure in the future. It connects the data network, linking the data provider and the data demander, and then implementing the data charge per time, by quantity, and by value, similar to the model of Didi and Uber.
In the future, privacy computing will realize the release of data value on a larger scale, continuously empowering data compliance interaction and data value conversion.
