Collaborative Research: CNS Core: Small: Privacy by Memory Design

差异隐私(DP)已被广泛接受为保护数据隐私的事实上的技术。尽管对概率分布的研究已经进行了十年，但仍然存在一个被忽视的关键研究问题，那就是所有现有的概率分布研究都是基于这样的假设，即软件可以方便而可靠地从概率分布中采样并添加噪声。然而，这一假设正不断受到最近关于其侵犯隐私的发现的挑战，以及可能缺乏高级软件库的低端设备对隐私保护的日益增长的需求。因此，本项目的创新研究角度是在现代电子设备中普遍存在的嵌入式存储器上直接实现DP机制。在技术方面，已开发的创新具有以下优点。(1)它将主机设备从专用软件中解放出来，实现了“设计隐私”的愿景；(2)它同时提高了能效、隐私和芯片开销等多种系统性能；(3)所开发的技术是原始的、通用的，可扩展到每一种电子设备。该项目还将对我们的社会、经济和劳动力发展产生深远的影响。具体地说，开发的技术将对许多敏感应用(例如监视和传感)和关键基础设施(例如物联网设备)产生变革。它可能会通过在芯片中添加隐私保护功能来增加美国芯片供应商的收入和竞争力，保护纳税人和企业的敏感数据，维护国家安全，并帮助美国在网络安全方面击败全球竞争对手。此外，该项目将帮助私人投资机构更新现有课程，吸引学生-优先考虑女性或/和第一代大学生-进行研究培训，并通过现有的成功项目，如NSF资助的阿拉巴马州莫比尔的RET网站，促进社区外展。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Differential privacy (DP) has been widely accepted as the de facto technique for protecting data privacy. Despite the decade-long research efforts on DP, there still exists a critical research problem that has been largely overlooked, that is all existing DP studies are grounded on the hypothesis that software can easily and faithfully sample and add noises from a probability distribution. However, this hypothesis is being constantly challenged by recent findings about its privacy violation and by the growing demand of privacy protection in low-end devices that may lack high-level software libraries. Hence, this project's innovative research angle is to realize DP mechanisms directly on embedded memories, which are ubiquitous in modern electronic devices. On the technical front, the developed innovation has the following merits. (1) It frees host devices from dedicated software and accomplishes the vision of "privacy by design"; (2) It concurrently improves manifold system performance such as power efficiency, privacy, and chip overhead; (3) The developed technique is primitive, generic, and scalable to every electronic device.This project will also create profound impact on our society, economy, and workforce development. Specifically, the developed technique will be transformative to numerous sensitive applications (e.g., surveillance and sensing) and critical infrastructures (e.g., Internet of Things devices). It will potentially increase the U.S. chip vendors' revenue and competitiveness by adding privacy-preserving functionality to their chips, protect taxpayers' and enterprises' sensitive data, safeguard national security, and help the U.S. out-compete global competitors in cybersecurity. Moreover, this project will help PIs to update existing curricula, engage students – with priority to female or/and first-generation college students – for research training, and promote community outreach through existing successful programs such as the NSF-funded RET site in Mobile, Alabama.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.