AdPriRec: A context-aware recommender system for user privacy in MANET services

Abstract

Mobile ad hoc network (MANET) has become a practical platform for pervasive services. Various user data could be requested for accessing such a service. However, it is normally difficult for a user to justify whether it is safe and proper to disclose personal data to others in different contexts. For solving this problem, we propose AdPriRec, a context-aware recommender system for preserving user privacy in MANET services. To support frequent changes of node pseudonyms in MANET, we develop a hybrid recommendation generation solution. We apply a trusted recommendation sever who knows the node's real identity to calculate a recommendation vector based on long term historical experiences. The vector can be also generated at each MANET node according to recent experiences accumulated based on node pseudonyms, while this vector could be further fine-tuned when the recommendation server is accessible. We design a number of algorithms for AdPriRec to generate context-aware recommendations for MANET users. The recommendation vector is calculated based on a number of factors such as data sharing behaviors and behavior correlation, service popularity and context, personal data type, community information of nodes and trust value of each involved party. An example based evaluation illustrates the usage and implication of the factors and shows AdPriRec's effectiveness. A prototype implementation based on Nokia N900 further proves the concept of AdPriRec design. © 2011 Springer-Verlag.