RFID and Databases
Guest: CSE Prof. Magdalena Balazinska
Magda works in the database field and is one of the major contributors to CSE's RFID Ecosystem project (which Evan discussed earlier).
Nearly all the privacy and surveillance concerns of the information age tie back somehow to fears of centralized databases that can be mined in some fashion. Magda will lead us in a discussion about how databases are used to obtain higher-order information in RFID deployments and some of the technical problems encountered. In particular, Magda wishes to discuss what happens where there are errors in the reading of tags or with sparse data. She will talk about techniques that her and others are using to deal with noisy data. How can historical data be used to smooth over inaccuracies? How can information about other people or objects be used to probabalistically infer what happened at a given questionable point?
Some questions to think about beforehand might be: (1) in what kinds of RFID deployments would data smoothing be useful/necessary? (2) how might data smoothing techniques be affected by data retention laws? (3) where is the smoothing taking place? in the data stored in the database itself or in the interface between the database and each application? and (4) if peers are using applications through which they might learn something about you, what might be some of the social consequences of inaccurate data smoothing?
- Read sections 1 and 3 of http://www.cs.washington.edu/homes/magda/mobide06.pdf which outlines some approaches to handling noisy data in databases
- Bring 2 or 3 questions to ask Magda about the capabilities of databases with respect to RFID