Access Type

Open Access Dissertation

Date of Award

January 2012

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Computer Science

First Advisor

Loren Schwiebert

Second Advisor

Weisong Shi

Abstract

With the wide spread of computer systems and networks, privacy has become an issue that increasingly attracts attention. In wireless sensor networks, the location of an event source may be subject to unintentional disclosure through traffic analysis by the attacker. In vehicular networks, authentication leaves a trail to tie a driver to a sequence of time and space coordinates. In a cloud-based navigation system, the location information of a sensitive itinerary is disclosed. Those scenarios have shown that privacy protection is a far-reaching problem that could span many different aspects of a computer/network system, especially on a diversified landscape of such systems.

To address privacy protection, we propose to look at the issue from three aspects. First, traffic analysis represents one class of problems. This is because in general, encryption can be applied to protect the information being transmitted but the pattern of transmission is harder to hide due to other constraints. To defeat traffic analysis, it is necessary to identify those constraints and decide the trade-off between them and privacy protection. We have shown that the threat to the source location privacy within a sensor network is directly related with the pattern of the routing protocol. Thus to completely remove that pattern, we propose to use a random walk to defeat the threat.

Second, authentication is generally required to establish the identities of interacting parties in an electronical communication. But it unnecessarily reveals other private information when it is applied to a vehicular network. We propose to introduce tunable anonymity, through both asymmetric and symmetric encryption primitives, into the authentication process so that it provides $k$-anonymity. We further extend the scenario to mobile scenarios.

At last, it is often perceived that private information has to be shared in order to obtain certain services. For example, source location and destination location have to be sent for looking up the shortest path between them. We show that it is possible to apply private information retrieval so that a service provider knows for whom it has provided service for accounting purposes, but not the details of the service.

In general, the three aspects represent some basic aspects of privacy issues arising from using computer/network systems. Our approaches, while innovative for the scenarios discussed at hand, are general enough to be applied to similar scenarios.

Share

COinS