CPS:SMALL: Privacy-preserving Network Congestion Control: Theory and Applications

该项目的目标是在实现可证明的隐私保证的同时，实现更大程度的人群感知数据共享。我们的方法是限制知识的位置轨迹的人群感知数据，以排除起源和目的地，然后考虑使用分布式控制和差分隐私技术的系统：这使得探索隐私成本的理论界限。所取得的成果将为构建和分析网络物理系统（CPS）创造一种新的方法，允许用户对人群感知数据做出决策，同时严格保护他们的隐私。该项目的重点外展活动，在开发一个软件环境，让学生探索隐私性能权衡在运输业务和先进的课程CPS的安全和隐私。在我们的技术方法，我们专注于用户输入的隐私，如起源（初始状态），目的地（偏好），和实用程序功能。这使我们能够建模和分析如何使用人群感知数据来推断这些敏感的输入，使用分布式控制和差异隐私领域的技术。隐私的各种概念将支持广泛的研究计划，包括专用网络控制的性能限制，不同类别的网络物理系统的设计原则，以及考虑位置流行度的新位置隐私度量。这些贡献推进了概率模型的形式化分析领域，以及控制和优化中新兴的隐私子领域。理论研究的动机和评估的模拟和概念验证的实现在群众来源的拥塞检测的背景下。

The goal of this project is to enable greater sharing of crowd-sensed data, while achieving provable privacy guarantees. Our approach is to limit knowledge of the location traces of crowd-sensed data to exclude the origin and destination, and then consider the system using techniques from distributed control and differential privacy: this enables exploration of the theoretical bounds on the cost of privacy. The achieved results will create a new approach for building and analyzing networked cyber-physical systems (CPS) that permits users to make decisions on crowd-sensed data, and at the same time protects their privacy in a rigorous sense. The project has focused outreach activities in developing a software environment for students to explore privacy-performance trade-offs in transportation operations and an advanced course on security and privacy of CPS.In our technical approach, we focus on the privacy of user inputs, such as the origin (initial state), destination (preference), and utility functions. This enables us to model and analyze how the crowd-sensed data can be used to infer these sensitive inputs, using techniques adapted from the fields of distributed control and differential privacy. The various notions of privacy will support a broad research program, including performance limits of private network control, design principles for different classes of cyber-physical systems, and new location privacy metrics that take into account location popularity. These contributions advance the field of formal analysis of probabilistic models, and the burgeoning subfield of privacy in control and optimization. The theoretical research will be motivated by and evaluated on simulations and proof-of-concept implementations in the context of crowd-sourced congestion detection.