国广清科：隐私计算技术在金融领域的探索与实践

随着《数据安全法》和《个人信息保护法》的施行，以及人们对隐私保护需求的日益增加，市场对数据安全保护意识逐渐增强。在数据要素市场化高度迫切、数据隐私与安全保护日益增强的今天，如何在保证数据安全、隐私合规的前提下，促进数据要素的有序流动与高效释放成为推动数据要素市场化配置的核心问题。隐私计算的系统化发展，有望成为技术最优解。

The market's awareness of data security protection is steadily growing as a result of the adoption of the Data Security Law and the Personal Information Protection Law, as well as the rising need for privacy protection. The most pressing problem in promoting the market allocation of data elements is how to promote the orderly flow and efficient release of data elements under the presumption of ensuring data security and privacy compliance. This is due to the highly urgent marketization of data elements and the increasing protection of data privacy and security. The best possible technology answer is anticipated to be the methodical development of privacy computing.

在数字经济浪潮下，尤其在当前金融业务形态更加开放的背景下，作为实现“数据不动价值动”的关键技术，隐私计算凭借保障数据安全、转化和释放数据价值等优势，正逐步发展为金融业的“刚需”。

Privacy computing is a key technology to achieve "data does not move the value of moving" in the wave of the digital economy, especially in the context of the current financial business form more open. With the benefits of protecting data, transforming it, and releasing the value of it, privacy computing is gradually developing into the financial industry's "just need" technology a "need" for the financial sector.

目前，金融业是隐私计算商业化应用最为成熟的行业场景之一。金融业对数据安全和隐私保护的要求更高、监管更严，这对隐私计算在金融场景中的落地起到极大推动作用。部分金融机构通过自研或与隐私计算厂商合作探索，正逐步使隐私计算在金融信贷风控、营销、反洗钱、反欺诈、保险定价与理赔、业务协同等场景下落地。

One of the most developed industrial scenarios for the commercialization of privacy computing is the finance sector at the moment. The use of privacy computing in financial situations is highly encouraged by the financial sector's higher standards and stronger regulations for data security and privacy protection. Through self-developed or collaborative exploration with privacy computing vendors, some financial institutions are gradually making privacy computing available in financial credit risk control, marketing, anti-money laundering, anti-fraud, insurance pricing and claims, business collaboration, and other scenarios.

银行作为传统金融机构的代表，在科技赋能的进化中，必然涉及到与外部数据的联合建模。银行也是隐私计算最可能率先完全落地的领域。首先，银行找到存量用户需补全画像标签，才能服务于流失召回、交叉营销场景，这非常依赖于银行外部的数据。而隐私计算中的匿踪查询可以保证银行在查询外部数据的时候，避免用户信息被缓存。并且，小微企业贷等对个人或者企业进行信贷评估的场景，也需要依赖外部数据源做联合建模评估。

As representatives of conventional financial institutions, banks must participate in collaborative modeling using outside data as technology empowerment progresses. Additionally, banks are the industry where privacy computing is most likely to be completely adopted initially.First of all, banks discover that the stock of users has to fill up the whole picture tag in order to support cross-marketing and churn recall scenarios, which heavily rely on data from outside the bank. When requesting external data, banks can avoid caching user information by using the anonymity query in privacy computing. In order to do collaborative modeling evaluation, scenarios like credit evaluation of people or businesses for micro- and small-business loans must also rely on external data sources.

国广清科作为一家专注于隐私计算技术研究和应用的数据技术服务公司，已经在金融领域积累了丰富的经验。在金融场景下，国广清科的隐私计算优势主要包括以下几个方面：

CRI TSING’S TECH, a provider of data technology services with an emphasis on privacy computing technology research and implementation, has a wealth of finance industry knowledge. CRI TSING’S TECH's privacy computing advantages primarily cover the following areas in the financial context:

1.技术架构

国广清科隐私计算的技术架构包括数据处理、数据加密、数据存储、数据传输等多个方面，保障了客户数据的高度安全性和保密性。同时，公司自主研发的“青稞”隐私计算一体机，在保障数据安全的同时，也提升了数据的可用性。“青稞”隐私计算一体机为了提高TEE内深度深度神经网络的训练速度，把深度神经网络的模型进行合理拆分，在TEE内保证训练数据的"可用不可见"，利用GPU对模型训练时性能敏感部分进行异构加速，从而显著的提高模型的训练速度，同时利用内置硬件增强隐私计算安全性和提升性能，为用户提供高性能、高安全的隐私计算基础设施，实现用户开箱即用、部署方便，能够降低项目综合启动成本并加速场景落地，使得金融机构可以更加灵活地处理数据，提高业务效率和精度。

1. Technical architecture

The high security and confidentiality of customer data is ensured by the technological architecture of privacy computing, which comprises data processing, data encryption, data storage, data transfer, and other elements. The company's in-house created "Barley" privacy computing all-in-one device can simultaneously ensure data confidentiality and increase data availability. The "Barley" privacy computing all-in-one machine divides the deep neural network model into manageable pieces, makes sure the training data is "available but not visible" in TEE, and uses GPU to speed up the performance-sensitive portion of the model training, significantly increasing the training speed of the model.

In addition, the built-in hardware is used to improve privacy computing security and performance, giving users access to a high-performance, high-security privacy computing infrastructure that is simple to deploy right out of the box, lowers overall project start-up costs, and speeds scenario implementation. This allows financial institutions to process data more adaptably and boosts business accuracy and efficiency.

2.数据安全保障

国广清科隐私计算采用了多层数据安全保障措施，包括数据加密、数据备份、访问控制等，确保了客户数据的绝对安全性。此外，公司还通过了中国信通院联邦学习专项评测和多方安全计算专项评测，确保了数据安全保障的权威性和可靠性。

2. Data security guarantee

To guarantee complete protection of client data, State Wide Clear Computing has implemented multi-layer data security measures, including data encryption, data backup, and access control. Additionally, the business has successfully passed the multi-party secure computing special evaluation as well as the special assessment of the China ICT Academy for Federal Learning, confirming the legitimacy and dependability of the data security guarantee.

3.业务流程优化

国广清科隐私计算可以帮助金融机构优化业务流程，提高效率和精度。例如，在贷款审批方面，公司可以通过隐私计算技术对客户的历史信用记录和其他数据进行分析，从而更快地做出决策。此外，在风险控制方面，公司可以通过隐私计算技术对客户的行为数据进行分析，从而更准确地预测风险。

3. Business Process Optimization

CRI TSING’S TECH Financial companies may enhance efficiency and accuracy by optimizing business operations with privacy computing. For instance, businesses may employ privacy computing technologies to examine past credit histories and other data to approve loans more quickly. Additionally, businesses may use privacy computing technologies to evaluate consumer behavioral data for risk control in order to more precisely identify dangers.

总的来说，国广清科的隐私计算解决方案在金融领域具有广泛的应用前景。通过采用多种技术手段，隐私计算可以帮助金融机构更好地保护客户数据隐私，同时提高数据处理的效率和准确性。同时，隐私计算还可以帮助金融机构优化业务流程，提高业务效率和准确性。因此，我们相信，在未来的发展中，隐私计算将会成为金融领域的重要趋势。

Overall, there are many potential applications for CRI TSING’S TECH's privacy computing solution in the financial industry. Financial firms may better secure consumer data privacy by using a variety of technical tools, and privacy computing can help by increasing data processing speed and accuracy. At the same time, privacy computing may assist financial firms in streamlining operations and raising productivity and accuracy levels. Therefore, we think that privacy computing will play a significant role in the financial industry going forward.