September 22, 2005

T3: 9/22 Can Bloom Filters power private personalization?


This week at 4PM, Yang Wang, a CommerceNet intern, will be reporting on his investigation into trading off personalization accuracy against privacy of users’ profiles:

Bloom Filter (BF), a space-efficient data structure, has been widely used in supporting membership queries, at the cost of certain false positive rate. In our attempt to utilize the false positives to protect users privacy in recommender systems, we compare the usage of BF and random obfuscation in a recommender system MultiLens, which runs a personalized movie recommendation website, MovieLens.