Yuval Elovici

This paper presents a new privacy model for hiding the information interests of a homogenous group of users who share a local area network and an access point to the Web. The suggested model is aimed at preventing eavesdroppers from using identifiable members' tracks to infer the group common interests (referred to as the group profile) while allowing members of the group to identify themselves to various services. The model consists of generating faked transactions in various fields of interest in order to confuse eavesdroppers by preventing them from accurate derivation of the group profile. A privacy measure is defined as one that reflects the degree of confusion a system can cause eavesdroppers. A prototype was developed implementing an HTTP proxy that derives the group profile and generates faked transaction in order to hide it. Experiments were conducted in order to examine the effectiveness of the model. Initial results are encouraging.