Web User De-Identification in Personalization

摘要

It is a kind of privacy infraction in personalized web service if the user profile submitted to one web site transferred to another site without user permission. That can cause the second web site easily re-identify to whom these personal data belong, no matter whether the transfer is under control or by hacking. This paper presents a portable solution for users to bind their sensitive web data under the appointed domain. Such data, including query logs, user accounts, click stream etc, could be used to identify the sensitive information of the particular user. By our domain stretching deidentification method, if personal data leak from domain A to B, the web user could still not be identified even though he logins to sites under domain B using the same name and password. In the experiment implemented by javascript, we show the flexibility and efficiency of our deidentification approach.