T-79.514 Special Course on Cryptology (2-6 cr)
Autumn 2003: Privacy-Preserving Data Mining
New:Webtopi registering opened (03.09.03)
The goal of data-mining is usually quite opposite to the privacy: different companies are interested in obtaining as much information about you and your friends as possible, to be able to use it later as they need --- either by storing everything, or just storing the necessary bits of the data that is necessary to later build up some models on the data. Clearly, people are however not interested in giving away their personal data for free, and thus might object in submitting their data at all.
In privacy-preserving data-mining, one encourages people to submit some data that helps in building up data models without revealing too much information about clients. If done properly, this might result in companies getting more data, and thus in them being able to build more appropriate models. PPDM also looks on other areas of data-mining, that are as important: e.g., on possibilities of running data-mining algorithms (e.g., ID3) on two separate databases, owned by different companies (e.g., the genome databases).
The PPDM area is currently booming, but many questions have been proven to be too difficult to solve. Thus, researchers have proposed many protocols that are not extremely secure, but still efficient.
The seminar is hopefully going to be a melting-pot: we expect both students with background in cryptography and security (from TCS, TML, ...) and in data-mining (CIS, University of Helsinki, ...). Due to that, there will be no formal prerequisites. However, to be sure, drop an email to Helger Lipmaa as soon as possible.
Hopefully, it is going to be a research seminar, at the end of which students (and teachers) will have at least enough ideas to start writing a couple of papers on the subject. In particular, one of the goal is to cryptanalyse existing papers. Practice has shown that most of the papers in this area either solve a wrong question (from the DM viewpoint), or are cryptographically weak. To improve on that, it would be essential to get students with a good data-mining background.
Want to know something about subject?
See the background page.
This page: http://www.tcs.hut.fi/Studies/T-79.514/
[TCS main] [Contact Info] [Personnel] [Research] [Publications] [Software] [Studies] [News Archive] [Links]
Latest update: 24 August 2004.