Date of Award
1-1-2022
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
College/School/Department
Department of Information Science
Content Description
1 online resource (vii, 86 pages) : illustrations (some color)
Dissertation/Thesis Chair
Sanjay Goel
Committee Members
Yuan Hong, Brian H Nussbaum
Keywords
Data Fraud Mitigation, Data Poisoning, Data Pre-clearance, Differential Privacy, SMC, Training Data, Data protection, Computer security, Data encryption (Computer science), Computer networks, Privacy, Right of
Subject Categories
Library and Information Science
Abstract
The growth of cloud-based services collecting user data for online analytical processing (OLAP), machine learning, and applications relating to the Internet of Things (IoT) has also increased concern with data privacy. Privacy-preserving data sharing using Secure Multiparty Communication (SMC) enables the exchange of encrypted data between parties to perform calculations while maintaining data privacy for both parties simultaneously. This research builds on established work that proposed a privacy-preserving driving style recognition protocol designed to work well with semi-honest actors. In this protocol, both parties follow the established algorithms and do not overtly attempt to deceive the other party into revealing data or subvert the evaluation mechanism. SMC is further leveraged through additional enhancements to the privacy-preserving driving style classification protocol, which detects several data tampering efforts. These new enhancements counter attempts to subvert the driving style classification process and generate a more favorable rating. In this research, a pair of cheating detection algorithms are proposed. The algorithms are scripted in the Python programming language, using open libraries for encryption, data aggregation, and analytics. For testing, multiple synthetic data sets based on elements of several publicly available data sets have been created, along with a template for creating additional data sets. The experimental results demonstrate the effectiveness and sensitivity of the new cheating- detection algorithms. The scripts, data sets, and data set generation template are available for independent verification and reproducibility.
Recommended Citation
Sprissler, Ethan, "Cheating detection in a privacy preserving driving style recognition protocol" (2022). Legacy Theses & Dissertations (2009 - 2024). 3027.
https://scholarsarchive.library.albany.edu/legacy-etd/3027