埃森哲数据实验室通过隐私保护计算(Privacy Preserving Computation)赋能洞察

发布时间: 2021-04-06

共享数据几乎成为了当下企业主们公认的重要需求，企业间的数据合作为企业发展创造了新的优势和机会，来自整个行业生态的综合数据提供了任何一家企业都无法单独挖掘的洞察和价值。

但即使企业急于释放综合数据的价值，他们也希望保持对自己数据的控制。在某些情况下，这并不只是管理者的私人偏好，因为原始数据往往过于敏感，无法共享。尽管行业内企业合作不断加深，但这两种基于数据安全与隐私的担忧都使企业无法从共享数据中获益。但这正是隐私保护计算（PPC）的用武之地。

在埃森哲实验室的区块链和分布式账本的研究中，Kirby Linvill领导了关于PPC的研究工作。开发了一个演示，展示了这些方法在金融行业的可行性，为暗池启用隐私保护技术创造了可能。

“暗池”听起来可能有些奇怪，但它只是一个私人证券交易所的术语。如果你是一个机构投资者，想完成一宗金额庞大的股票交易，显然易见你不会希望在交易之前，有关交易的消息泄露出去。如果暗池以外的公开市场听说你打算出售一家公司的100万股股票，那么在交易完成之前，价格可能会有重大变化。但如果你在暗池中发布自己的交易意向，就找不到合适的买家或卖家。可一旦发布了交易意向，却又面临信息泄露导致交易失败的风险。

我们已经研发了一种方法，让机构投资者在没有信息风险的情况下参与暗池交易。利用我们的合作伙伴开发的最新PPC应用，我们可以在不公开数据的情况下匹配暗池中各方的交易意向与需求。这意味，暗池里的投资者不需要向其他参与者披露机密信息，甚至不需要向暗池运营商披露。事实上，投资者看不到任何其他参与者的信息；他们只有在自己的订单或需求得到匹配时才会得到通知。这可以让暗池高效、高交易量地运作，而不存在机密信息泄露的风险，从而防止市场操纵。

本演示的技术核心是一种特殊的PPC技术，称为安全多方计算（MPC）。MPC允许多方对私人数据进行联合计算，而不会将私人数据透露给任何其他方。竞争企业可以对其共享数据执行互惠互利的分析，同时确保自己的敏感原始数据永远不会被任何其他企业使用。这一技术在除金融以外的许多领域都很有价值，例如，一些国家政府可以使用MPC对敏感的人口普查数据进行集体分析，而不需要与其他国家共享原始的基础数据。

MPC只是KirbyLinvill和我正在与实验室研发团队一起探索的革命性PPC技术之一。我们也在探索同态加密和企业可用的安全处理器。同态加密允许在加密数据上进行计算，而无需先解密（或者根本不需要解密）。在硬件方面，安全处理器是特殊模块，允许在硬件内进行数据处理，加密的专用存储器区域直接在微处理器芯片上。

PPC技术对跨行业的公司来说尤为有用。无论是银行、医疗保健还是制造业，敏感信息的隐私保护都是关键，但数据协作有着重要的价值。安全数据共享，是实现最大化协作的前提，我们相信也会是当下行业最热门的技术议题。

原文：

Companies today know they have to share data to succeed. Enterprise partnerships create new advantages and opportunities for growth, and the combined data from across the ecosystem offers insights and value that’s impossible to uncover alone.

But even as companies are eager to unlock the power of pooled data, they’d like to maintain control over their own. In some cases it’s more than just a preference; the raw data may actually too sensitive to share. Both of these concerns have kept companies from reaping the benefits of shared insights even as ecosystem partnerships have grown. But that’s where Privacy-Preserving Computation (PPC) comes in.

I drive research on blockchain and distributed ledgers at Accenture Labs, and Kirby Linvill has led the Labs’ PPC research efforts. We’ve worked together to develop a demonstration of what’s possible with these approaches for the financial industry, enabling privacy-preserving technology for dark pools.

“Dark pool” may sound ominous, but it’s just a term for a private securities exchange. If you’re an institutional investor that wants to make a large stock trade, you don’t want news about that trade to leak out before you make it. If the public market hears that you’re looking to sell a million shares of a company, the price could change significantly before your trade is complete. But you can’t find a buyer for stock you want to sell – or a seller for stock you want to buy – without telling the dark pool that you’re interested. And the second you tell people you’re interested in buying or selling, you create the risk of a leak.

We’ve demonstrated a way to let institutional investors participate in dark pools without that risk. Leveraging the latest PPC advancements from some of our partners, we can blindly match offsetting trading positions from different partners in the pool. What does that mean for dark pool investors? They don’t need to disclose confidential information to other market participants, or even to a centralized dark pool operator. In fact, they can’t see the positions or instruments of any other participant; they’re only informed when there’s a match for their current orders. This could allow dark pools to operate efficiently and at high volume, without the risk of leaked confidential information – preventing market manipulation.

The technical heart of this demonstration is a particular PPC technique called secure Multi-Party Computation (MPC). MPC allows multiple parties to run joint computations on private data without revealing that private data to any other party. Competing enterprises can perform mutually beneficial analytics on their shared data, while ensuring their own sensitive raw data is never usable by any of the other companies in the group. These same approaches could be valuable in many spaces, not just finance; you can imagine nations using MPC to run collective analytics on sensitive census data without needing to share the raw underlying data with other nations, for example.

MPC is only one of the revolutionary PPC techniques that Kirby Linvill and I are exploring with our Labs R&D teams. We’re also exploring Homomorphic Encryption and enterprise use of secure enclaves. Homomorphic encryption allows computation on encrypted data without the need to decrypt it first (or at all). On the hardware wide, secure enclaves are special modules that allow for data processing within hardware-provided, encrypted private memory areas directly on the microprocessor chip.

Privacy-preserving computational techniques offer major value for companies across industries. Whether it’s banking, healthcare, or manufacturing, the privacy of sensitive information is key, but there is significant value to be gained from data collaboration.

本文内容转载自：埃森哲数据实验室  https://www.accenture.com

原文作者：Giuseppe Giordano

原文地址：https://www.accenture.com/us-en/blogs/technology-innovation/giordano-linvill-privacy-preserving-computation

作者: Giuseppe Giordano