Verida 技术 Litepaper:自主机密计算网络以确保私人 AI 的安全(第2部分).

Sure, here is the HTML structure with the translated text in simplified Chinese: ```html

这是接下来三周内发布的三篇文章中的第二篇,旨在发布《Verida技术轻论文:自主可控的保密计算网络,保护私密人工智能》。文章最初由Verida的CEO兼联合创始人Chris Were发布。第一部分请看这里

``` This HTML structure maintains the original format while displaying the translated text in simplified Chinese.

保密计算

越来越多的大型云服务提供商提供访问可信执行环境(TEE)的保密计算服务。这些包括:AWS Nitro,Google Confidential Compute 和 Azure Confidential Compute。近期还出现了像 Marlin Oyster 和 Super Protocol 这样的令牌化保密计算服务。

这些计算方案通常允许在安全TEE硬件上的安全飞地内部部署容器(如Docker实例)。 飞地具有一系列验证和安全措施,可以证明在飞地内部运行的代码和数据是您所期望的代码,并且飞地已经以一种抗篡改的方式部署。

在这些安全隔离区域中存在一些重要的限制,即:

  1. 没有可从基础设施运营商直接访问飞地的途径。通信是通过安全飞地和主机之间的专用虚拟插口进行的。
  2. Certainly! Here's the translation of the English text "There is no disk storage available, everything must be stored in RAM" into simplified Chinese, while maintaining the HTML structure: ```html

    没有可用的磁盘存储空间,一切都必须存储在RAM中。

    ``` In this HTML snippet: - `

    ` and `

    ` are paragraph tags to indicate the beginning and end of the paragraph. - The Chinese text inside the `

    ` tags is the translated sentence. This HTML structure ensures that the translated text is presented as a paragraph in the context of a web page or document.

  3. 直接GPU访问通常在安全区域内不可用(对于高性能LLM训练和推断是必要的),但预计这种能力将在2025年初提供。

Certainly! Here's the translated text in simplified Chinese, keeping the HTML structure intact: ```html

(*) 在某些情况下,基础设施运营商控制硬件认证密钥和云基础设施,这引入了安全风险,需要仔细解决,但超出本文档的范围。

```

Sure, here's the translated text in simplified Chinese while keeping the HTML structure intact: ```html

The Verida network effectively functions as a database providing high-performance data synchronization and decryption. While secure enclaves do not have local disk access (by design), it is possible to provide a secure enclave with a private key. This allows the enclave to quickly download user data, load it into memory, and perform operations.

``` Translated to simplified Chinese: ```html

Verida网络实际上是一个提供高性能数据同步和解密的数据库。虽然安全飞地设计上没有本地磁盘访问权限,但可以向安全飞地提供私钥,使其能够快速下载用户数据,加载到内存中并进行操作。

```

尽管飞地无法直接访问互联网,但可以通过在主机和飞地之间建立安全套接字连接来“代理”网络请求到外部世界。这会增加对飞地安全的可能攻击面,但也是与其他网络服务交互的机密计算所必需的条件。

```html

机密的AI推断对用户提示的快速响应时间至关重要,以确保终端用户的高质量体验。通过机密计算直接访问GPU很可能是满足这些要求所必需的。目前,带有可信执行环境(TEE)的GPU访问受到限制,然而像NVIDIA H100这样的产品提供了这些功能,并且这些功能将在适当的时候供Verida网络使用。

```

自主计算

Sure, here is the translated text in simplified Chinese, formatted in an HTML structure: ```html

Verida 提供一个基于机密计算基础设施的自主计算基础设施堆栈。

``` In this HTML snippet: - `

` denotes a paragraph tag, used to structure the text appropriately. - The Chinese text translates the original English sentence into simplified Chinese.

Figure 1: Self-Sovereign Compute Architecture

自主计算基础设施提供以下保证:

  1. 用户数据不可被基础设施节点运营者访问。
  2. 运行时代码可以被验证,以确保其运行了期望的代码。
  3. 用户完全控制其私人数据,并可以随时授予/撤销对第三方的访问权限。
  4. 第三方开发者可以构建和部署代码,以保密的方式操作用户数据。
  5. 用户完全控制可以在其数据上运行的计算服务,并可以随时授予/撤销对第三方的访问权限。

有两种不同类型的计算,具有不同的基础设施要求; 无状态保密计算和有状态保密计算。

无状态(通用)保密计算

Sure, here's the HTML structure with the translated text in simplified Chinese: ```html

这种计算是无状态的,它在API请求之间不保留任何用户数据。然而,它可以从其他API请求用户数据,并以保密的方式处理该用户数据。

``` In simplified Chinese, the translation of the text is: "这种计算是无状态的,它在API请求之间不保留任何用户数据。然而,它可以从其他API请求用户数据,并以保密的方式处理该用户数据。"

Sure, here's how you can represent that text in HTML with the translation: ```html

这里有一些关于通用无状态计算的示例,这些示例将在网络上运行。

``` This HTML structure keeps the translated text while maintaining the basic HTML formatting with a `

` (paragraph) tag.

Figure 2: Verida Personal Data Bridge

私人数据桥有助于用户连接到第三方平台的API(如:Meta,Google,Amazon等)。这些节点必须以保密方式运行,因为它们存储API机密,处理终端用户的访问/刷新令牌到第三方平台,从这些平台提取敏感用户数据,然后使用私人用户密钥将该数据存储在Verida网络上用户的私人数据库中。

LLM APIs接受包含敏感用户数据的用户提示,因此它们必须在保密的计算环境中运行。

Certainly! Here's the translated text in simplified Chinese, keeping the HTML structure: ```html

AI API(例如AI提示服务和AI代理服务)为用户数据和LLM之间的交互提供了“粘合剂”。 AI服务可以使用用户数据API(见下文)直接访问用户数据。这使得它能够通过LLM API促进检索增强生成(RAG),利用用户数据。由于请求的结果,这些API还可以将数据保存回用户的数据库(即将数据保存到向量数据库中,以供未来的RAG查询)。

``` This HTML snippet retains the structure while presenting the translated text in simplified Chinese.

保持HTML结构不变,将以下英文文本翻译成简体中文: 请参阅第1部分中的“自主AI交互模型”,了解这些通用计算服务如何相互交互,以提供用户数据上的AI服务。

Sure, here is the translation of "Stateful (User) Confidential Compute" in simplified Chinese while keeping the HTML structure intact: ```html 有状态(用户)保密计算 ```

这种计算是有状态的,用户数据会在内存中持续一段时间。这提高了性能,最终增强了终端用户的用户体验。

用户数据API将使授权的第三方应用程序(如私人AI代理)能够轻松快速地访问解密的私人用户数据。假设有一个单一的用户数据API,但实际上很可能会有多个在不同基础设施上运行的API服务。

以下是一些可以访问的数据类型示例:

  1. 在多个平台(Telegram、Signal、Slack、Whatsapp等)上的聊天记录
  2. 网络浏览器历史
  3. 企业知识库(例如:Notion,谷歌云端硬盘等)
  4. 电子邮件
  5. 金融交易
  6. 产品购买
  7. 健康数据

每种数据类型都有不同的容量和大小,这也会因用户而异。预计个人用户所需的总存储空间将在100MB至2GB之间,而企业知识库将会更大。

在第一阶段,重点将放在结构化数据上,而不是图片或视频。这与Verida现有的存储节点基础设施保持一致,为人工智能数据互操作性的首次迭代提供支持并协助数据模式的开发。

用户数据API公开了支持以下数据服务的端点:

  1. Sure, here's how you could structure the HTML while displaying the translated text in simplified Chinese: ```html Translate Text to Chinese

    Authentication for decentralized identities to connect their account to a User Data API Node

    去中心化身份验证,将他们的帐户连接到用户数据 API 节点。

    ``` In this example: - `lang="zh-CN"` sets the language of the document to simplified Chinese. - The `

    ` tag contains the translated text as the main heading. - The `

    ` tag provides the translation in a paragraph format.

  2. 获取第三方应用程序的访问和刷新令牌的身份验证
  3. Certainly! Here is the translation of "Database queries that execute over a user’s data" into simplified Chinese, while keeping the HTML structure: ```html 执行在用户数据上的数据库查询 ``` This HTML structure maintains the original text but in simplified Chinese.
  4. 关键词(Lucene)式搜索用户数据
  5. 矢量数据库搜索用户数据

连接有状态计算与去中心化身份。

第三方应用程序获得访问令牌,该令牌允许根据用户授予的同意,以有限的方式访问用户数据。

To translate the given English text into simplified Chinese while keeping the HTML structure intact, you can use the following: ```html

在Verida网络上的去中心化身份可以授权网络上的三个或更多自主计算节点,以管理第三方应用程序对其数据的访问。这是通过身份的DID文档中的serviceEndpoint功能实现的。这与当前的Verida数据库存储网络以分配存储节点来负责用户数据的方式相同。

``` Here's the translated text in simplified Chinese: "在Verida网络上,去中心化身份可以授权三个或更多的自主计算节点,以管理第三方应用程序对其数据的访问。这是通过身份的DID文档中的serviceEndpoint功能实现的。这与当前的Verida数据库存储网络分配存储节点来负责用户数据的方式相同。"

安全隔离区没有磁盘访问权限,但用户数据可在Verida网络上获取(经过加密),并可根据适当的用户私钥随需同步。在需要时,用户数据需要“热加载”,这涉及从Verida网络同步加密的用户数据,解密后存储在内存中,并添加其他元数据(如搜索索引)。当进行初始API请求时,这将发生,确保用户数据已准备好供第三方应用程序快速访问。

在一段时间的闲置后(例如,1小时),用户数据将会从内存中卸载,以节省底层计算节点的资源。这样,一个单独的用户数据API节点可以同时为多个去中心化身份提供服务。

为了最大程度地减少终端用户的第一个交互时间,确保“热加载”足够快是必要的。这些计算节点还必须具备足够的内存,以一次加载多个用户的数据。Verida开发了一个内部概念验证,以验证“热加载”概念与用户数据将是一个可行的解决方案。

Certainly! Here's the translation in simplified Chinese, embedded within an HTML structure: ```html

为了增强隐私和安全性,每个去中心化身份的数据和执行将在安全计算节点的保密计算隔离虚拟机中运行。

``` In this HTML snippet, `

` denotes a paragraph tag, commonly used for displaying text content on a webpage.

敬请期待,Litepaper的第三和最后发布将在下周发布。

2024-08-28 04:33:50 AI中文站翻译自原文