There is no doubt that privacy computing is on the eve of large-scale application. More and more privacy computing target projects have emerged, especially in recent times, which has grown rapidly, which to a certain extent represents the development of technology and market recognition.
Recently, Tianmian Technology won the bid for UnionPay Business' "Privacy Computing Software Platform Procurement Project", demonstrating its outstanding technical strength and scenario case advantages in the field of privacy computing. Taking this opportunity, PCview Privacy Computing Research Institute interviewed Xu Wenbin, technical director of privacy computing at Tianmian Technology, and talked about Tianmian Technology’s achievements and development in the field of privacy computing, as well as future plans.
Xu Wenbin introduced that Tianmian Technology began to enter privacy computing in 2018. Since customers did not know that they were willing to accept it, the hardest thing for doing it early was to persevere. However, because it has more than 50 million users, Tianmian Technology has been implemented in real scenarios from the beginning, and its products are relatively mature and available. This is Tianmian Technology's natural advantage in implementing financial scenario applications.
He said that Tianmian Technology’s next layout planning for privacy computing technology can be divided into three stages. The first is to combine scenarios to create practical privacy computing technology solutions; the second is to open up various systems to achieve docking and integration, and truly promote the development of enterprise business; and finally, to build a more complete data ecosystem to completely break the data silo and achieve Interconnection promotes the healthy and orderly development of the industry.
1. Emphasis on new technologies, entering privacy computing in 2018
WeLab Huili Group is a well-known financial technology group in Asia, providing diversified financial technology services, including operating Asia's first licensed virtual bank - WeLabBank (Huili Bank) and other purely online consumer financial services in China. It enjoys a high reputation in the Hong Kong, Mainland China and Indonesia markets, with more than 50 million users and more than 700 corporate customers.
Tianmian Information Technology (Shenzhen) Co., Ltd. (hereinafter referred to as Tianmian Technology) is a subsidiary of WeLab Huili Group. Xu Wenbin said that WeLab Huili Group began to expand its mainland business in 2014. Because it has caught the express train of mobile Internet finance, its business has developed rapidly. Since then, Tianmian Technology has become an independent entity as the unified external output window of WeLab Huili Group's financial technology.
Xu Wenbin studied computer science at university and has been engaged in big data development for a long time after graduation. Early data can be obtained in plain text, which obviously involves great risks. Therefore, Tianmian Technology began to lay out the field of privacy computing as early as 2018. He introduced that his department is the Data Intelligence Laboratory. The entire team is composed of experts from the fields of data, risk control, and marketing. There are also experts related to cryptography and certification agencies. This is very consistent with privacy computing.
Xu Wenbin also said that the data intelligence laboratory where he works is also affiliated with the Welab Huili Group Innovation Research Center. "The purpose of establishing the Innovation Research Center is to understand new technologies and solve practical problems. We have been able to develop because we have grasped the technical point of big data risk control, so we attach great importance to technological innovation and application."
2. Position ToB services and output products around core technologies
Tianmian Technology is positioned as a technology company that provides ToB services and outputs around technology themes. The products are various systems centered on specific businesses, such as credit and risk control businesses, including: incoming parts system, channel management system, certification platform, anti-fraud system, intelligent decision-making system, and intelligent collection system. The other is to launch general-purpose systems, including data middle platforms, privacy computing, and AI laboratories. These are not only used in the financial field, but are also common in other industries.
Just looking at privacy computing, up to now, Tianmian Federal Learning Platform has established cooperation with dozens of financial and traditional institutions. The content of the cooperation is mainly to unite data providers, and jointly conduct joint operations on the premise that the data of each party does not leave the private domain. Risk control and joint marketing model training.
Typical case: A leading mutual financial company needs to carry out activation marketing for existing silent users. It can only achieve break-even by using only existing own data features for modeling or only referring to third-party data scores. However, offline third-party data is introduced for this purpose. The joint modeling method also has the risk of user data leakage; by using the Tianmian Federated Learning Platform, third-party data is introduced on the basis of its own data characteristics for joint modeling, which effectively improves the model effect and marketing ROI. Mutual gold company's marketing revenue increased by 680,000 in the current period.
3. With a user base of over 50 million, financial applications have natural advantages.
The financial industry was the first to be exposed to privacy computing. Tianmian Technology relies on WeLab Huili Group's business accumulation over the years and currently has more than 50 million users. For these users, whether they are doing risk control or marketing, the technical requirements are actually very high. Because there is such a sample, it is of great help to product development. By implementing it in real scenarios and conducting verification, Tianmian Technology can discover problems in time, such as algorithm, function, interface problems, etc., and then make adjustments and improve them to the outside world. Output, the products created in this way are relatively more mature and usable. This is also Tianmian Technology’s natural advantage in implementing financial applications.
Xu Wenbin said that the most difficult thing about implementing privacy computing is the early promotion. At that time, many customers did not understand it and needed to explain what federated learning is, what multi-party secure computing is, and what its uses are. Now users have begun to discuss in which scenarios privacy computing is useful, in which departments, and how to use it with better results. Therefore, he concluded that if privacy computing is done early, the difficulty is to persist. If you want to do privacy computing well, the focus is still on paying attention to customer needs. Paying attention to how users use privacy computing to bring maximum benefits under compliance principles is the purpose of ToB customer service.
4. Future planning: building a more complete data ecosystem
At present, privacy computing has moved from tools to applications. Tianmian Technology has already had relevant application cases in marketing and risk control, and some medical application cases have also been observed externally. Xu Wenbin said that these implementation cases also prove the changes in privacy computing over the years. Now more and more companies are paying attention to privacy computing, and more and more scenarios are beginning to be implemented.
He believes that privacy computing will become a basic technical facility and can be integrated with more and more scenarios. Take marketing as an example. In the past, people used to bring external data to the local, whether it was modeling or application, it was basically a clear text transmission. Some would use some MD5, SHA256 and the like for alignment, but now privacy intersection, Stealth query, federated learning may be used in the intermediate joint modeling, which is more compliant and meets actual needs.
The "2021 Privacy Computing Industry Research Report" released by KPMG shows that the domestic privacy computing market is ushering in a period of rapid development. It is expected that technical service revenue will reach 10 billion to 20 billion yuan in three years, and is even expected to leverage hundreds of billions of data platforms Operating income space.
Xu Wenbin believes that as big as the market for data transactions is, the market for privacy computing should be as big. The cooperation method can be a buying and selling system, a profit sharing method, or a consulting service method. The original model is not suitable and some adjustments can be made. The development space is quite worth looking forward to.
Xu Wenbin introduced that in terms of privacy computing, Tianmian Technology’s future planning can be roughly divided into three stages, as follows:
First, implement practical privacy computing technology solutions. With data as the core and combined with specific scenarios, we develop various privacy computing systems that comply with regulations and data applications.
Secondly, by opening up various systems and achieving integration and integration, we can truly promote business development. Most of the privacy computing project of Tianmian Technology is open source, and federated learning is completely open source. Tianmian Technology hopes to provide a de facto standard for the industry, bring convenience to the industry, and allow everyone to use it; at the same time, through combination with other scenarios, it will also broaden the boundaries of the entire privacy computing. From not knowing it at the beginning to now understanding and using it, the convenience of open source will help privacy computing to be implemented faster, which will have an excellent promotion effect on both the industry and Tianmian Technology.
Finally, build a more complete data ecosystem based on privacy computing. Because technology is only the underlying foundation, if we want this technology to be better accepted, we must break the so-called data islands and string the data together into an interconnected state. There are many privacy computing companies now, but the problem has not been solved. The original data island has become a data archipelago. Therefore, Tianmian Technology will also participate in the formulation of some standards in the future to open up the archipelago through industry standards, thereby building a more complete data ecosystem.