Cognitive Privacy for Personal Clouds

Abstract

This paper proposes a novel Cognitive Privacy (CogPriv) framework that improves privacy of data sharing between Personal Clouds for different application types and across heterogeneous networks. Depending on the behaviour of neighbouring network nodes, their estimated privacy levels, resource availability, and social network connectivity, each Personal Cloud may decide to use different transmission network for different types of data and privacy requirements. CogPriv is fully distributed, uses complex graph contacts analytics and multiple implicit novel heuristics, and combines these with smart probing to identify presence and behaviour of privacy compromising nodes in the network. Based on sensed local context and through cooperation with remote nodes in the network, CogPriv is able to transparently and on-the-fly change the network in order to avoid transmissions when privacy may be compromised. We show that CogPriv achieves higher end-to-end privacy levels compared to both noncognitive cellular network communication and state-of-the-art strategies based on privacy-aware adaptive social mobile networks routing for a range of experiment scenarios based on real-world user and network traces. CogPriv is able to adapt to varying network connectivity and maintain high quality of service while managing to keep low data exposure for a wide range of privacy leakage levels in the infrastructure.