Elastic pathing

Abstract

Today, people have the opportunity to opt-in to usage-based automotive insurances for reduced premiums by allowing companies to monitor their driving behavior. Several companies claim to measure only speed data to preserve privacy. With our elastic pathing algorithm, we show that drivers can be tracked by merely collecting their speed data and knowing their home location, which insurance companies do, with an accuracy that constitutes privacy intrusion. To demonstrate the algorithm's real-world applicability, we evaluated its performance with datasets from central New Jersey and Seattle, Washington, representing suburban and urban areas. Our algorithm predicted destinations with error within 250 meters for 14% traces and within 500 meters for 24% traces in the New Jersey dataset (254 traces). For the Seattle dataset (691 traces), we similarly predicted destinations with error within 250 and 500 meters for 13% and 26% of the traces respectively. Our work shows that these insurance schemes enable a substantial breach of privacy.

Lessons Learned: Surveying the Practicality of Differential Privacy in the Industry

“…Unlike traditional privacy techniques, which are vulnerable to auxiliary information attacks [8,29,33,67,79,94], differential privacy [28] mathematically formalizes a privacy guarantee agnostic to background information. A function guarantees differential privacy (e.g., an analytics query or a machine learning (ML) model) if it bounds the information gain that an attacker can expect from its outputs.…”

“…Attacks on privacy (Q2) could enable malicious analysts to misuse data for spying on or leaking secret information of celebrities, acquaintances, "friends", or relatives [100], blackmailing, or discriminating individuals in social or commercial transactions online [35]. Despite the risks, companies predominantly use traditional and potentially vulnerable anonymization techniques (e.g., pseudo-anonymization or k-anonymity), as demonstrated by the research community [8,29,33,67,79,94]. Thus, we suggest companies increase efforts to research and deploy more advanced PETs.…”

“…Several factors have spurred the development of more advanced privacy-enhancing technologies (PETs) in the past years. On the one hand, from an adversarial perspective, (i) multiple white-hat attacks have shown that "traditional" anonymization techniques such as suppressing names are vulnerable to re-identification across industries [8,29,33,67,79,94]. Additionally, between 2020 and 2021, (ii) the total cost of data breaches have increased by 10% on average [87].…”

Context-Aware Security for Vehicles and Fleets: A Survey

“…Less sensitive are data about the environmental conditions or the technical condition of the vehicle [162]. Since even a speed signal is sufficient to trace PII such as location [163], the transmission of raw data from the vehicle to a backend to establish behavior monitoring is very limited. For processing and storage, especially outside the vehicle, either explicit authorization must be granted or legal regulations must require it.…”

On my way: Optimizing driving routes for navigation applications

“…Recently, Chang et al [19] designed and implemented an Android-based navigation app, which can help reduce the traveling time, fuel consumption and emitted carbon dioxide. Gao et al [20] have developed an "elastic pathing" algorithm, which can be used to derive driving routes merely with the knowledge of the starting location and time-stamped speed data.…”