LP-SBA-XACML. (c2019)

Abstract

The wide applicability of Internet of Things (IoT) would truly enable the pervasiveness of smart devices for sensing data. IoT coupled with machine learning would enter us in an era of smart and personalized, services. In order to achieve service personalization, there is a need to collect sensitive data about the users. That yields to privacy concerns due to the possibility of abusing the data or having attackers to gain unauthorized access. Moreover, the nature of IoT devices, being resource and computationally constrained, makes it di cult to perform heavy protection mechanisms. Despite the presence of several solutions for protecting user privacy, they were not created for the purpose of running on small devices at a large scale. On top of that, existing solutions lack the customization of user privacy in which users have little to no control over their own private data. In this regards, we address the aforementioned issue of protecting user's privacy while taking into account e ciency as well as memory usage. The proposed scheme embeds an e cient and lightweight algebra based that targets user privacy and provides e cient policy evaluation. Moreover, an intelligent model to customize user's privacy based on real time behavior is integrated. Experiments conducted on synthetic and real-life scenarios to demonstrate the feasibility and relevance of our proposed framework within IoT environment.