Securing Data Storage

计算机用户通过各种存储介质（SD卡、u盘、笔记本电脑、电子邮件附件等）存储大量的个人数据和知识产权，而这些用户没有任何存储安全保证。有些技术，如加密，在某些地方被用于保护传输，但它们是存储数据的一个方面，而不是与数据相关的策略。有许多改进存储的技术，例如加密、安全删除、可否认存储、历史独立性和存储虚拟化。尽管如此，它们通常不会一起使用，也没有一个在实践中被广泛使用。虽然这对于某些用户来说不是最优的，但对于在敌对环境中工作的极端用户，如军队、记者和政治异见者，它完全失败了。通过将安全性作为标准文件系统的一部分提供，这项工作将提高存储安全性的最低标准。该研究项目有两大组成部分。其中一个重点是提高便携式闪存存储的安全性，作为第一道防线。闪存被广泛使用，但它有一个细微的特点：数据不能被覆盖。相反，新数据被写错了位置，删除只能在大粒度下发生。这对存储、读取和删除数据的算法和文件系统有重要的影响。它还使一些通常很容易提供的安全功能变得非常复杂。基于对flash文件系统设计挑战的深刻理解，本工作将设计下一代flash文件系统，其目标是协调加密，安全删除，可否认性和其他功能，而不会在最终文件系统的可用性和性能方面造成重大损失。第二个组成部分是开发可用的技术，允许用户为数据指定存储策略，而不必指定要使用的特定存储介质。目前，用户把文件放在u盘上便于携带，然后通过电子邮件发送给自己，以便在其他电脑上访问。用户经常进行临时备份，并将数据存储在多个位置，而不清楚如何存储数据，甚至不清楚数据存储在哪里。目前人们用来保护数据的技术不方便、不安全、不可靠或不可用。我们将创建虚拟存储设备（VSD）来解决这些问题。VSD将成为数据进出存储端点的路由器，允许用户专注于他们想要的存储属性，如安全性、速度、可靠性和成本。VSD将从对用户有意义的策略和风险属性中抽象出存储介质的物理属性（例如，它们位于何处以及如何连接）。随着数据存储与物理计算设备分离，用户将无处不在地访问、处理和计算他们的数据。

Computer users store immense amounts of personal data and intellectualproperty across a variety of storage media (SD cards, USB sticks, laptops,attachments in e-mails, etc.), and these users are not given any guaranteesabout storage security. Some techniques, like encryption, are used in someplaces or to secure transmissions, but they are an aspect of what stores thedata and not a policy tied to the data.There exists a variety of techniques for improving storage, such as encryption, securedeletion, deniable storage, history independence, and storage virtualization.Despite that, they are frequently not used together and none are widely used inpractice. While this is sub-optimal for some users, it completely fails the world'smore-extreme users working in hostile environments, such as military, journalists,and political dissidents. This work will raise the minimum standard for storage securityby providing security as part of standard file systems.The research program has two broad components. One focuses on improving thesecurity of portable flash-based storage to act as a first line of defense.Flash memory is widely used but it has a characteristic nuance: data cannot beoverwritten. Instead new data is written out of place, and deletion can onlyhappen at a large granularity. This has significant implications on thealgorithms and file systems that store, read, and delete data. It also severelycomplicates several security features that would normally be easilyprovided. Building on a deep understanding of the challenges in designing flashfile systems, this work will design a next-generation flash file system with thegoals of harmonizing encryption, secure deletion, deniability, and other featureswithout a significant cost in the usability and performance of the resulting filesystem.The second component is developing usable techniques to allow users to specifystorage policy for data without having to specify the particular storage mediumto use. Currently, users put files on USB sticks to make them portable, ore-mail it to themselves to access it on other computers. Users frequently makead hoc backups, and store data in multiple places without a clear understandingof how, or even where, the data is stored. Current techniques people use tosafeguard their data are inconvenient, insecure, unreliable, or unusable. We willcreate a Virtual Storage Device (VSD) to address these problems. The VSD willbe a router for data to and from storage endpoints, allowing users to focus onthe storage properties that they desire, such as security, speed, reliability, andcost. The VSD will abstract the physical properties of storage media (e.g., wherethey are located and how they are connected) from the policy and risk attributesthat are meaningful to the user. Users will access, work on, and compute withtheir data ubiquitously as the data storage is separated from the physicalcomputation devices.