Defending Malicious Check-In Using Big Data Analysis of Indoor Positioning System: An Access Point Selection Approach

The integration of WiFi fingerprint-based indoor positioning technology and big data analysis emerges as a new research prospect. Through the analysis of big data collected from users' submission, we can discover many other applications of fingerprint positioning. A popular application is the check-in to point of interest (POI) for its crowd traffic evaluation according to the volume of received signal strength (RSS) fingerprints submitted by users. However, this crowd traffic evaluation method may be susceptible to the intrusion of malicious check-in behaviors. Attackers who are not at the target POI submit the self-modification RSS fingerprints that can be located at the target POI in order to illegally increase its crowd traffic. To this end, we propose a malicious check-in defense scheme based on the access point (AP) selection to resist attackers who aim to successfully initiate the fingerprint modification. Specifically, the distance between different POIs in fingerprint space is firstly developed for AP selection. Then, in order to increase the robustness of selected AP subset, we propose the mutual information among different classes as a selection condition. Through the multiobjective optimization and Pareto optimality, we can obtain the best AP subset to participate in the computation of positioning algorithm. Furthermore, the optimal modified fingerprint is searched by the level set method (LSM), which can be utilized to measure the costs of attackers and the robustness of the system. In addition, we propose an iterative weight updating method based on classification error to learn the optimal weight in order to balance the positioning accuracy and robustness. We finally carry out extensive simulations to validate that the POI crowd traffic can be assessed in terms of the RSS fingerprint-related information and our proposed scheme can perform high robustness to resist malicious check-in.