Date of Award

12-2011

Level of Access

Campus-Only Thesis

Degree Name

Doctor of Philosophy (PhD)

Department

Spatial Information Science and Engineering

Advisor

Max J. Egenhofer

Second Committee Member

Per Erik Gårder

Third Committee Member

Lars Kulik

Additional Committee Members

Reinhard Moratz

Michael F. Worboys

Abstract

Location tracking through mobile devices has become a major challenge to privacy and security. While location tracking may be regulated by laws and controlled through the setting of a mobile device such as a cell phone, the technical way of dealing with it— by processing the user information—may provide protections in case the first two measures fail.

Current studies on user information processing are mainly found in two directions: anonymization and obfuscation. Anonymization mixes the movements of a group of users or masks a user with one or more pseudonyms, while obfuscation processes the location information without altering a user's identity. Anonymization generally does not survive the analyzing of the visited places. Processed locations, on the other hand, are prone to cracking through data-mining algorithms on a user's spatiotemporal trajectory.

This thesis develops a way of representing and modeling location information such that conversions can be made between footprints—regions on maps determined by geographic coordinates—and placenames. Then privacy may be preserved for each location by yielding a placename that often cannot be traced uniquely and precisely back to the original footprint.

The research introduces an integrated framework, with which a set of entities is represented by a structure of its numerical (quantitative) measurements and a structure of its verbal (qualitative) descriptions. A measurement or description can be transformed to a designated level of coarseness within the same structure, and as well converted to a corresponding description or measurement in the other structure7 8he framework concretizes into an integrated approach for location information through simplified gazetteer models, which are established from structural representations of geographic entities in the geometric space and the name space, the quantitative and the qualitative aspects of geographic space, respectively.

Through the integrated approach, the encoding of a unique, precise footprint yields a placename that is not necessarily unique, imprecise, and possibly ambiguous, while the decoding of a placename does a reverse processing. The approach shows flexible applicability to real-world situations through a demonstration with a region in Maine, providing a platform for modeling strategies toward location privacy to be developed.

Share