Distinguished Lecturer Series

When Less is More: Using Data Fidelity to Protect Privacy and Save Energy

Carla Ellis

Duke University

Friday, February 24, 2006
10:30am - SENSQ 5317

Refreshments at 10:00am

Hosted by José Brustoloni

Abstract

Mobile computing devices with access to wireless networking and inexpensive sensors offer the promise of many exciting new context-aware applications. Unfortunately, the promise of ubiquitous context-aware applications has not been realized as rapidly as hoped. Two major barriers to the widespread adoption of this technology have been (1) limited battery lifetime of the mobile devices carried by users or the remotely deployed sensor nodes instrumenting the environment and (2) perceived threats to privacy by what might be viewed as surveillance.

While there have been great advancements in the design of low power hardware and in power-aware computer architecture and system software, developers of applications for energy-constrained devices can exploit knowledge of the actual user requirements. Choosing an appropriate level of data fidelity is one approach available to application developers to conserve energy consumption. Based on experience with context-aware systems, it appears that data fidelity also can play a role in safeguarding user privacy.

In this talk, I will describe a model of data fidelity that captures the tradeoffs related to energy consumption and privacy in context-aware applications. The model identifies dimensions of data fidelity such as the capture, the persistence, and the dissemination of sensor and context data. I will illustrate the model in terms of several energy-aware and privacy-sensitive projects. Finally, I will describe elements of the Context-Awareness Fidelity Expression (CAFÉ) development toolkit that is intended to guide application developers toward a fidelity level that is consistent with the actual requirements of the application and thus avoid exposing users to unnecessary privacy risks or the energy consumption overhead of working with excessive data.

Biography of Speaker

Carla Ellis received the B.S. degree from the University of Toledo, Toledo OH, in 1972 and the M.S. and Ph.D. degrees from the University of Washington, Seattle, in 1977 and 1979. Before coming to Duke as an Associate Professor in 1986, she was a member of the Computer Science faculties at the University of Oregon, Eugene, from 1978 to 1980, and at the University of Rochester, Rochester NY, from 1980 to 1986. During the 1997-98 academic year, she was on sabbatical at the University of Washington. She is an Experimental Computer Scientist with broad interests in operating systems for mobile, parallel, and distributed systems. Some of her current interests and projects include dynamic data-driven applications of sensor networks, the Milly Watt Project for energy management in computing systems, and privacy in ubiquitous, context-aware systems.

Professor Ellis is the Editor-in-Chief of ACM Transactions on Computer Systems (TOCS), a member of the board of the Computing Research Association (CRA) and currently its Secretary, co-chair of the Academic Alliance of the National Center of Women and Information Technology (NCWIT) and a member and former co-chair of CRA's Committee on the Status of Women in Computing Research (CRA-W). She has also formerly served as Chair of ACM SIGOPS, as a member of the ACM Council and Executive Committee, and as Program Chair of the USENIX Annual Technical Conference.