CPS: Medium: A Unified Framework for IoT Privacy

智能城市、联网车辆、智能家居和联网医疗设备都是物联网（IoT）如何通过利用丰富的用户特定数据来显著改善用户体验，从而在未来几十年内彻底改变我们生活的例子。 然而，共享这些数据可能会损害用户的隐私，这种对用户隐私的威胁已经成为物联网广泛采用的关键风险。 这凸显了对网络物理系统科学至关重要的一个重要而根本的挑战：即使物联网数据被仔细匿名化，由于生成的数据量庞大，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化的方法来理解物联网系统中隐私的基本基础，并开发可证明的私有物联网实现，这些实现对模型中的不确定性具有鲁棒性。 这种方法的一个关键优点是它可以实现可证明的隐私，即，没有算法可以破坏用户的隐私。 该项目还继续了团队的教育和对不同学生的参与，包括本科生在研究项目中的重要参与，并创建和推广免费和开放获取的教育材料。该项目中考虑的技术问题分为两个主要方向。 在Thrust 1中，构建了物联网隐私的理论基础。 主要目标是获得对物联网方法的实用性可以保持的程度的基本理解，同时采用隐私保护机制，以通过将给定的跟踪与先前的用户行为相匹配来可证明地防止对手损害用户的隐私。 这一推动力的关键是实现稳健和模型独立的设计，即，以最少的关于系统和数据模型的假设集实现完美的隐私。在Thrust 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Thrust 1的结果应用于联网车辆应用。 更一般地说，这将表明给定用户的数据可以在多大程度上对感兴趣的对手保持私密，同时仍然支持使用此类服务。

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.