In order to promote the excellent practice of privacy computing technology in the financial industry, the Privacy Computing Alliance (PPCA) and the China Communications Standards Association Big Data Technology Standard Promotion Committee (CCSA TC601) jointly held the "Privacy Computing Financial Anti-Fraud Application Salon" on September 15, 2022. "Conference, Huang Miao, general manager of Impulse Online Solutions, was invited to participate in this event and delivered a speech entitled "Anti-fraud Application of Privacy Computing Combining Software and Hardware".
#1 Trusted Execution Environment TEE - International Mainstream Solution
As the only feasible technical means to ensure the safe circulation of data, privacy computing has three technical routes: cryptography, federated learning, and trusted execution environment. Among them, the trusted execution environment combined with hardware security can ensure that a task is executed as expected, and ensure the confidentiality and integrity of the initial state and runtime code. Intel SGX, TrustZone, and AMD SEV are currently the most popular implementation methods.
As the mainstream TEE technology, community and ecology in foreign countries are becoming more and more mature, localized chip manufacturers have also begun to make efforts in the direction of TEE, and the formation of an independent and controllable complete community and ecology in my country is just around the corner. Even manufacturers that have seen heterogeneous computing hardware GPUs this year have begun to introduce TEE functions, and Nvidia's H100 will be the first GPU to support TEE.
Except for TrustZone, the various TEE solutions currently in use basically appeared after 2010 or even 2015. In just six or seven years, the mainstream chip manufacturers who proposed TEE have released at least two or even three generations of major version upgrades of the TEE core architecture. It not only reflects from the side that TEE technology is more and more accepted by the market, but also shows that the market is also continuously bringing demand for TEE hardware and software.
#2 CPU&GPU Privacy Computing Solution
Trusted Execution Environment (TEE) technology and GPU have natural integration advantages. Using GPU to speed up TEE is the best solution for privacy computing technology to reduce performance loss at present and in the future.
In the sharing, Huang Miao introduced the most advanced domestic privacy computing solution combining CPU and GPU proposed by Impulse Online, which extends the security and privacy capabilities of the trusted execution environment to GPU computing chips to build trusted computing power. Supports models that process more data volumes and data dimensions, and supports processing machine learning and deep learning models, greatly improving computing efficiency and lowering the threshold for application promotion.
#3 The industry's first high-performance heterogeneous privacy computing all-in-one computer
The industry's first domestically-made high-performance heterogeneous privacy computing all-in-one machine designed and released by Impulse Online in conjunction with chip manufacturers. At the hardware level, domestically-made TEE can directly connect to domestically-made GPUs, so that AI models can use GPU accelerator cards and ensure the running accuracy safety.
At the software level, build an open and shared platform for Impulse AI based on heterogeneous trusted computing power, provide credible and safe AI services, ensure the security of data and models to be tested, and enable financial institutions to have safe AI service opening capabilities. For example, in the context of open banking, financial institutions will export maturely applied AI capabilities (such as OCR capabilities, NLP capabilities, etc.) The data security of the provider, revitalizing the industry model assets to realize the external operation and value realization of AI assets.
At present, the Xinchuang high-performance heterogeneous privacy computing all-in-one machine built by Impulse Online has completed large-scale implementation in multiple scenarios such as the Institute of Information and Communication of the Ministry of Industry and Information Technology, China Telecom, and Wuhan Yuntong, and has been unanimously recognized by many customers.