Do online social network friends still threaten my privacy?

Abstract

A user's online social network (OSN) friends commonly share information on their OSN profiles that might also characterize the user him-/herself. Therefore, OSN friends are potentially jeopardizing users' privacy. Previous studies demonstrated that third parties can potentially infer personally identifiable information (PII) based on information shared by users' OSN friends if sufficient information is accessible. However, when considering how privacy settings have been adjusted since then, it is unclear which attributes can still be predicted this way. In this paper, we present an empirical study on PII of Facebook users and their friends. We show that certain pieces of PII can easily be inferred. In contrast, other attributes are rarely made publicly available and/or correlate too little so that not enough information is revealed for intruding user privacy. For this study, we analyzed more than 1.2 million OSN profiles in a compliant manner to investigate the privacy risk due to attribute prediction by third parties. The data shown in this paper provides the basis for acting in a risk aware fashion in OSNs.

Anonymization Techniques for Privacy Preserving Data Publishing: A Comprehensive Survey

“…Fang et al [185], Labitzke et al [186], Youyang et al [187] 7. spatial and temporal information of the SN use fusing temporal and spatial data with other related information such as check-in/posts data to infer private information of users.…”

Multi-Dimensional Conversation Analysis Across Online Social Networks

“…7 We have put distances into 100 mile buckets. 8 Locale is the user chosen interface language for Facebook.com, such as EN US (English in USA), EN GB (English in Great Britain) In the previous sections, we have shown how different dimensions of user generated texts affect how many times a particular tweet/comment will be liked or commented. Regardless of these numbers, our geolocation experiments show that users who like/comment a text are more likely to live closer to the owner of the text.…”

Privacy Inference Attack Against Users in Online Social Networks: A Literature Review

“…The study [53] inferred user's attributes by emotionoriented mining and predicted user's attributes by construct an association graph based on public friends. F A Zamal et al evaluated the inference accuracy obtained by using features in Twitter profile and friend posts to enhance user attributes, showed that different subsamples of user neighbourhood depict different aspects of users themselves [54].…”