Privacy-enhanced middleware for location-based sub-community discovery in implicit social groups

Ahmed Elmesiry, Seungmin Rho*, Dmitri Botvich

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

11 Citations (Scopus)


In our connected world, recommender services have become widely known for their ability to provide expert and personalize information to participants of diverse applications. The excessive growth of social networks, a new kind of services are being embraced which are termed as “group based recommendation services”, where recommender services can be utilized to discover sub-communities within implicit social groups and provide referrals to new participants in order to join various sub-communities of other participants who share similar preferences or interests. Nevertheless, protecting participants’ privacy in recommendation services is a quite crucial aspect which might prevent participants from exchanging their own data with these services, which in turn detain the accuracy of the generated referrals. So in order to gain accurate referrals, recommendation services should have the ability to discover previously unknown sub-communities from different social groups in a way to preserve privacy of participants in each group. In this paper, we present a middleware that runs on end-users’ mobile phones to sanitize their profiles’ data when released for generating referrals, such that computation of referrals continues over the sanitized version of their profiles’ data. The proposed middleware is equipped with cryptography protocols to facilitate private discovery of sub-communities from the sanitized version of participants’ profiles in a university scenario. Location data are added to participants’ profiles to improve the awareness of surrounding sub-communities, so the offered referrals can be filtered based on adjacent locations for participant’s location. We performed a number of different experiments to test the efficiency and accuracy of our protocols. We also developed a formal model for the tradeoff between privacy level and accuracy of referrals. As supported by the experiments, the sub-communities were correctly identified with good accuracy and an acceptable privacy level.

Original languageEnglish
Pages (from-to)247-274
Number of pages28
JournalJournal of Supercomputing
Issue number1
Early online date10 Dec 2015
Publication statusPublished - 1 Jan 2016
Externally publishedYes


  • Clustering
  • Community recommendations
  • Middleware
  • Privacy
  • Secure multiparty communication


Dive into the research topics of 'Privacy-enhanced middleware for location-based sub-community discovery in implicit social groups'. Together they form a unique fingerprint.

Cite this