Tencent cloud security and privacy computing has passed the evaluation of the ICT Institute and obtained national recognition

2021 year 6 month 24 Japan , Held by China information and Communication Research Institute “2021 Big data industry summit - Results Conference ” Held in Beijing , At the press conference, the 12th batch of Xintong Hospital “ Big data product capability evaluation ” The results are officially announced , Tencent cloud security and privacy computing has become the first batch of products in China that have passed the special test of multi-party security computing performance and the special test of federal learning performance , continuity 3 In, it was recognized by the national authority .

Special evaluation certificate of multi-party security computing performance

Special evaluation certificate of federal learning performance

The privacy Computing Alliance of China Academy of communications has long focused on the field of privacy computing , Its privacy Computing Series evaluation is the earliest in the industry 、 Largest scale 、 The most comprehensive coverage evaluation . Including multi-party secure computing 、 Federal learning 、 Trusted execution environment 、 Evaluation of four functions such as blockchain assisted privacy computing and multi-party security computing 、 Federation learning two performance measures . The standards on which the test is based are the industry formed through the discussion of major experts in the industry / Group standards , The test process has been verified by principle 、 Algorithm validation 、 Function demonstration 、 Performance testing 、 Expert evaluation and other links , With high scientificity and preciseness . At present, it has become one of the most authoritative third-party evaluation brands in the field of privacy computing , The products evaluated represent the highest level of privacy computing power in China .

This year, the Chinese Academy of ICT conducted the first batch of special multi-party security computing performance tests and federal learning performance tests . Due to the security of privacy computing products 、 Accuracy and time-consuming restrict each other , Therefore, it is necessary to pass the security 、 After the accuracy test , Time consuming achievements are effective . Security includes communication security 、 Identity Authentication 、 The result is safe 、 Password security and other general security tests , It also includes the algorithm security verification for each type of algorithm . During this evaluation, Tencent cloud security and privacy computing is in the security 、 The evaluation of accuracy is also stable at the top level in the industry .

The performance test of multi-party secure computing and the performance test of Federated learning products focus on different algorithms , How to implement the product 、 The framework and security requirements are also different . The first multi-party secure computing performance tests include basic operations 、 Joint statistics 、 Hidden query 、 Safety is the key 、 Feature Engineering 、 Joint forecasting 、 Algorithm categories such as joint modeling , Tencent cloud security privacy computing is based on basic computing 、 Joint statistics 、 Feature Engineering 、 The performance of model training and reasoning is significantly better than the average level . The first batch of federal learning product performance tests include safe delivery 、 Feature Engineering 、 Joint modeling 、 Algorithm categories such as joint prediction , Tencent cloud security and privacy computing in XGB Model training and reasoning 、 Feature engineering has the absolute advantage to obtain the best performance evaluation performance .

among , We are concerned about anti fraud 、 Pre loan application score 、 The algorithms commonly used in financial scenarios such as crowd redirection have been optimized , Make the gradient lifting tree algorithm （XGB） And Feature Engineering （IV/WOE） Training time is greatly reduced , It is further convenient for customers to repeatedly adjust parameters during model training . in addition , For massive data application scenarios , Products based on Angel PowerFL Framework for distributed computing , The measured performance is better than the average value 8 More than times , The best performance evaluation performance has been achieved in the participating products .

Because of Finance 、 Government affairs and other scenarios affect joint risk control 、 Joint statistics 、 Strong statement of hidden query , At the beginning of this year, we upgraded aegis federal computing to Tencent cloud security and privacy Computing , The product is based on Tencent Angel PowerFL Privacy computing framework , Learn from the Federation （FL）、 secure multi-party computation （MPC）、 Trusted execution environment （TEE） Based on privacy data protection technology , For machine learning 、 Data analysis and other algorithms for customized privacy protection transformation , Create a privacy computing product with full data application link . The upgraded Tencent cloud security privacy computing product ensures that the privacy computing task can be completed quickly without the original data going out of the local area , While ensuring data security, it can give full play to the maximum value of data , It solves the problem of data island faced by enterprises . At present, Tencent cloud security and privacy computing platform is in the bank 、 insurance 、 government affairs 、 Education and other industries have benchmarking cases , Its excellent safety 、 Performance is reflected in many risk control systems 、 They all play an important role in the marketing scenario , That is to ensure data security and maximize the value of enterprise data .