Intelligent comprehensive privacy protection system for location-based services

Abstract

Recently, location-based systems (LBS) have been proven to be an essential element of smart cities due to the valuable benefits they provide to users searching for their nearest Points of Interest (PoI), facilitating daily life activities. However, privacy protection is a major concern in LBS, where attackers can apply advanced attacks, such as location homogeneity, semantic location and query analyzing attacks, to infer sensitive information about the private lives of LBS users. Therefore, protection of location privacy as well as query privacy is necessary to increase the trust of users in LBS. To address this issue, we present the Intelligent Comprehensive Privacy Protection (IntCPP) system as an enhancement of our previous work by employing a deep-learning technique. The Foursquare weekly trajectory dataset is selected to train the proposed system using the long short-term memory (LSTM) technique with an efficient pre-processing stage to adopt time-series data to the environment of LSTM. Evidence of the IntCPP system’s superiority is provided through comparison to two intelligent dummy-based systems as well as three traditional dummy-based systems. In terms of accuracy, a (0.05) enhancement degree is achieved, while in terms of entropy, cumulative resistance against attacks, and average cumulative cache hit ratio, (2.0, 100%, 0.17) enhancement degrees are achieved, respectively.