Session: Mobility Data Modeling: Values, Security, and Privacy
Chair: Jeff Ban
Cluster: Optimization Applications (Communication, Energy, Health, ML, ...)
Talk 1: Evaluation of data value in a transportation network
Speaker: Yueyue Fan
Abstract: Major transportation network operators, such as the California Department of Transportation, depend on good quality data to operate and manage their systems. Data could come from various sources, such as internally-owned sensors that collect traffic state information including speed, density, and volume, as well as third-party vendors that acquire, package, and sell mobile sensing data for road networks. A practical question would be: What data should the network operator acquire to best support planning and operations of a transportation network? In this talk, we will discuss the conception and modeling of the problem for understanding the value of data in the context of missing data imputation, and we will showcase its applicability using a case study based on PeMS data.
Talk 2: Data Poisoning Attack and Defense in Intelligent Transportation Systems
Speaker: Jeff Ban
Abstract: As data is becoming ubiquitous in intelligent transportation systems (ITS), data poisoning attacks emerge as a new threat. This research discusses the specific features of data poisoning attacks in ITS and proposes sensitivity-based optimization models for data poisoning. Lipschitz-based analysis methods are developed, with applications on well-studied ITS applications. Insights on how to defense data poisoning attacks are also presented with a few case studies.
Talk 3: Efficient Privacy-Preserved Processing of Multimodal Data for Vehicular Traffic Analysis
Speaker: Meisam Mohammady
Abstract: We estimate vehicular traffic states from multi- modal data collected by single-loop detectors while preserving the privacy of the individual vehicles contributing to the data. To this end, we propose a novel hybrid differential privacy (DP) approach that utilizes minimal randomization to preserve privacy by taking advantage of the relevant traffic state dynamics and the concept of DP sensitivity. Through theoretical analysis and experiments with real-world data, we show that the proposed approach significantly outperforms the related baseline non- private and private approaches in terms of accuracy and privacy preservation.
