CPS: Medium: A Unified Framework for IoT Privacy

智能城市、互联车辆、智能家居和互联医疗设备都是物联网(IoT)有望在未来几十年通过利用大量特定于用户的数据显著改善用户体验，从而彻底改变我们的生活的例子。然而，共享此类数据可能会危及用户的隐私，这种对用户隐私的威胁已经成为物联网广泛采用的一个重大风险。这突显了对网络物理系统科学至关重要的一个重要且根本性的挑战：即使物联网数据经过仔细的匿名化，由于生成的数据量巨大，以及对手使用强大的数学技术将当前行为与过去用户行为的痕迹相匹配，也可能发生重大的隐私泄露。该项目将开发一种系统的方法来了解物联网系统中隐私的基本基础，并开发可证明为私有的物联网实施，以稳健地应对模型中的不确定性。这种方法的一个关键优点是它可以实现可证明的隐私，即任何算法都不能破坏用户的隐私。该项目还继续了该团队对不同学生的教育和参与，包括本科生对研究计划的重要参与，并创建和推广免费和开放获取的教育材料。项目中考虑的技术问题被组织为两个主要方面。在推力1中，构建了物联网隐私的理论基础。主要目标是从根本上了解物联网方法的效用可以在多大程度上得到维护，同时使用隐私保护机制来通过将给定的跟踪与先前的用户行为进行匹配来可证明地防止对手损害用户的隐私。这一推力的关键是实现稳健和独立于模型的设计，即在对系统和数据模型的假设集最少的情况下实现完美隐私。在推力2中，为了验证理论并展示该方法的潜在影响，该项目利用团队的领域专业知识将推力1的结果应用于联网车辆应用。更广泛地说，这将表明在多大程度上可以对感兴趣的对手保密给定用户的数据，同时仍支持使用此类服务。

Smart cities, connected vehicles, smart homes, and connected healthcare devices are examples of how the Internet of Things (IoT) are expected to revolutionize our lives in the decades ahead by exploiting a wealth of user-specific data to significantly improve user experiences. However, sharing of such data can compromise a user's privacy, and this threat to user-privacy has emerged as a critical risk to the widespread adoption of IoT. This highlights an important and fundamental challenge critical to the Science of Cyber-Physical Systems: even if IoT data is carefully anonymized, significant privacy leaks can occur due to the sheer amount of the data generated and the use of powerful mathematical techniques by an adversary to match current behavior with traces of past user behavior. This project will develop a systematic approach to understand the fundamental underpinnings of privacy in IoT systems, and develop provably private IoT implementations that are robust to uncertainties in the models. A key advantage of this approach is that it can achieve provable privacy, i.e., no algorithm can break the privacy of the user. The project also continues the team's education and engagement of a diverse set of students, including the significant involvement of undergraduate students in the research program, and creates and promotes free and open access educational materials.The technical problems considered in the project are organized into two main thrusts. In Thrust 1, the theoretical foundations for IoT privacy are built. The main goal is to obtain a fundamental understanding of the degree to which the utility of IoT approaches can be maintained while employing privacy-preserving mechanisms to provably prevent an adversary from compromising a user's privacy by matching a given trace to prior user behavior. Critical to this thrust is achieving robust and model independent design, i.e., achieving perfect privacy with the minimum sets of assumptions about the system and data models. In Thrust 2, to validate the theory and demonstrate the potential impact of the approach, the project leverages the domain expertise of the team to apply the results of Thrust 1 in connected vehicle applications. More generally, this will indicate the degree to which the data of a given user can be kept private from an interested adversary while still supporting the use of such services.