On Taming the Inference Threat in Social Networks

This paper reports on the privacy threat posed by inferences in Online Social Netoworks. Specifically, it discusses the difficulties involved in taming this important privacy threat and describes, at a high-level, the design of a mechanism to tackle this problem. This mechanism is hybrid, combining privacy and transparency enhancing technologies. 1 The Inference Threat The vast reach of Web 2.0 services, in particular Online Social Networks (OSNs), has led to enormous accumulations of private user data in the hands of Web 2.0 service providers. For example, Facebook has over 900 millions monthly active users sharing private information on the service’s website [9]. This user data is highly important and valuable for OSN providers. That is because they follow a data-centric business model, offering most or all of their services for free and generating the lion’s part of their revenue by selling advertising space. For example, in 2011 Facebook had a revenue of US$3.154 billion only from selling advertising space on its website [8]. Heavily data-centered business models are also seen in, among others, Google, Twitter and LinkedIn. The price for targeted advertising space depends primarily on how well the advertisements fit the recipients’ interests and purchasing power. Depending on a particular threshold, a certain advertisement will or will not be displayed to the user. Therefore, detailed user profiles are essential for OSN providers’ success. These user profiles are created using the different types of data users implicitly and explicitly disclose to the provider. Schneier identifies six types of user data in OSNs [17]: • Service data: Data a user discloses to join an OSN. • Disclosed data: Data a user discloses on the own OSN page (e.g. a picture). • Entrusted data: Data a user discloses on other users’ pages (e.g. a post). • Incidental data: Data other users disclose about a user (e.g. tags on pictures). • Behavioral data: Data about a user’s behavior in the OSN (e.g. user’s IP or articles read). • Derived data: Data about a user that is derived from all other data types (e.g. user’s location inferred from the IP address). Service data, disclosed data and entrusted data are data that an OSN user actively discloses with the first being necessary to join the OSN and the other two being results of the user’s own active disclosure behavior within the OSN. Controls, offered both by OSN providers and