Unlock the value of data fusion! Tencent angel powerfl won the "leading scientific and Technological Achievement Award" at the 2021 digital Expo

Reading guide / Introduction

5 month 26 Japan -28 Japan , stay 2021 China international big data industry expo On , With the grasp of cutting-edge trends and technological leadership , Tencent big data - On the Tiangong platform Angel PowerFL Secure joint computing technology , Won the “ Leading technology achievement award —— New technology ” prize .

The digital Expo is the world's first Expo with the theme of big data , By the national development and Reform Commission 、 Ministry of industry and information technology 、 The state Internet information office and the people's Government of Guizhou Province jointly sponsored . As a digital Fair “ highlight ”, “ Leading technology achievement award ” So far, it is the only professional award with the name of the Expo and the theme of big data recorded by the national science and Technology Award Office .

As the fourth algebra intelligence fusion computing platform developed by Tencent “ Tencent big data - Tiangong ” Our first product ,Angel PowerFL The secure joint computing platform focuses on the cutting-edge technology of privacy protection computing , It can be used without disclosing data privacy , Realize multi-party data cooperation in a secure way , Solve the current widespread problem of data islands , Further release data value . Currently based Angel Power FL, Tencent security on Tencent cloud 、 Aegis and other products are opening the advantage of Tencent's big data privacy computing power to users .

2021 Leading technology achievement award -“ New technology ”： tencent AngelPowerFL Secure joint computing platform

 

Tencent big data Angel PowerFL Chengyong, the person in charge of the platform, said :“ It is a great honor to be recognized by the digital Expo ,Angel PowerFL from 2019 Started R & D in , Deeply cultivate core technologies , Some breakthroughs have been made in platform security and performance , While protecting data privacy , Help data collaboration . It's already in Finance 、 advertisement 、 Landing in business scenarios such as government affairs , Good application results have been achieved .”

Take the financial risk control scenario for example , A large financial service institution is based on Angel PowerFL platform , When the original data is not out of the local , A credit risk control model is constructed by combining multiple data sources . Compared with the traditional one-sided data modeling , be based on Angel PowerFL Build a federated model , stay AUC and KS The indicators are 5% and 16% The promotion of ( As shown in Figure 1 )——AUC And KS It is a common index to judge the accuracy of the model , The greater the numerical , The more accurate . While improving the accuracy of the model , It also meets the more stringent data privacy requirements in the financial scenario .

Credit risk control model - Comparison between unilateral modeling and federated modeling

besides , stay A bank Of Consumer finance business scenarios Next , also through too Angel PowerFL Platform operation The federal Study , Use multi-party data federation To build the One Risk control model type . stay PoC Stage , use 100 ten thousand Test data Next ,KS Indicators from Unilateral data modeling 0.4 Upgrade to 0.6, get 50% The promotion of （ Here's the picture ）, The effect of Federated modeling is remarkable .

In the video recommendation scenario , Traditionally , Because of the video APP There was no history browsing when the new user of logged in for the first time 、 Play record , Therefore, the user portrait cannot be generated , It is impossible to achieve accurate video recommendation . be based on Angel PowerFL platform , Use the other side ( Or user portrait library ) Sample label data of the user stored in , Can help APP Realize intelligent recommendation for new users , Help solve new users “ Cold start ” problem . At the same time, the user data cannot be exported locally , The privacy is fully guaranteed .

Here's the picture , In the real world ,Angel PowerFL Platform help this video APP Of The retention rate of new users is 21.82% The promotion of , The average playing time of new users has increased 3.47%, And effectively solve the recall problem of new users , It brings obvious cost optimization .

The federated modeling effect in the video recommendation scene

 

In the above three actual scenarios , be based on Angel PowerFL Secure joint computing platform , By building a multi-party data joint computing capability to ensure data privacy and security , On the one hand, it has greatly improved the effect of the model on the existing basis ; On the other hand , By breaking the data island 、 Realize data collaboration , It has brought new room for improvement to the business system .

With cloud computing 、 The application and development of new generation information technologies such as big data and artificial intelligence , Mining value from massive data 、 Stimulate the vitality of the data element market , It has become the main goal of the digital economy era .

therefore , For data users , What matters is how to balance the relationship between data fusion and privacy protection , It can realize deeper value mining through data collaboration , At the same time, it can fully ensure data privacy and security 、 Meet compliance requirements , Boost confidence of data owners —— Finally, data fusion and data privacy are realized “ Cooperative game ”, Accelerate the development of digital economy , Become the core challenge faced by all enterprises .

Privacy computing paves the way for this vision , This also makes it one of the hottest technology fields in the industry . As a multi domain 、 Interdisciplinary technical system , Privacy computing includes federated learning 、 secure multi-party computation 、 Trusted computing, etc , Involving artificial intelligence 、 Data Science 、 cryptography 、 Security protection technology and other technical fields .

Angel PowerFL Secure joint computing platform , It can provide users with leading privacy computing power . First, based on decentralized architecture , It can effectively avoid single point security risk ,3072 Bit high intensity homomorphic encryption calculation , Far beyond the of financial regulation 2048 Bit encryption level requirements ; secondly , Because the bottom layer adopts asynchronous and concurrent mechanism , At the same time, the encryption operation is optimized at the instruction level ,Angel PowerFL Performance beyond the industry 5 times , The distributed architecture adopted by the platform , Support the calculation of 100 billion scale ; Besides , Support joint data modeling and joint data analysis , Provide end-to-end full scene privacy computing capability and one click cloud native deployment capability .

Angel PowerFL Technical architecture

Based on big data 、 machine learning 、 Distributed 、 R & D experience and technical accumulation in security encryption and other technical fields , Tencent big data Angel PowerFL The team has been widely recognized by the industry , And won many awards .

Including in the international top privacy computing competition iDASH 2020 in ,Angel PowerFL The team won the championship with excellent results , The technical strength has been fully confirmed .

iDASH 2020 award certificate

Besides ,Angel PowerFL The team has also actively participated in the formulation of a number of standards related to the privacy computing industry , He has published many papers at world-class academic conferences , It has submitted relevant technical patents 20 Pieces of .

future ,Angel PowerFL The team will continue to forge ahead , In the field of privacy computing , Improve platform capability , Provide assistance for data collaboration in more business scenarios , To be safe 、 The way of compliance , Unleash the value of data fusion .