Towards Privacy-Preserving Networked Autonomous Mobility: Analysis, Tools Development, and Real-World Evaluation

The development of public autonomous mobility, including self-driving vehicles and delivery robots, which benefits from data-driven machine learning and deep learning algorithms, requires urgent investigation into the privacy issue. To protect the sensitive information in the collected data, the learning process should be studied to preserve privacy and also to measure privacy leakage. This project aims to investigate: (1) existing works about differential privacy theory and privacy-revealing techniques; (2) how to effectively apply privacy-preserving methods to autonomous mobility tasks in order to achieve privacy-preserving under the differential privacy framework; and (3) the privacy risk of using private datasets in the training process. To this end, the authors aim to design a privacy-preserving method that can be used in normal autonomous mobility settings and design an attack method to verify how much sensitive information can empirically be revealed during the training process.

Language

  • English

Media Info

  • Media Type: Digital/other
  • Edition: Final Report
  • Features: References;
  • Pagination: 3p

Subject/Index Terms

Filing Info

  • Accession Number: 01889956
  • Record Type: Publication
  • Contract Numbers: 69A3551747111
  • Files: UTC, NTL, TRIS, ATRI, USDOT
  • Created Date: Aug 14 2023 8:52AM