iLoc: Non-invasive Localization for Mobile Devices with COTS WiFi Access Points

In this work, we present iLoc, a non-invasive localization system for mobile devices with COTS WiFi access points (APs). We designed and implemented this working system, which only uses COTS APs to locate mobile devices without hardware nor software modifications. Moreover, iLoc is user-friendly because it need not the initial location of the users and need not any settings relating to it. iLoc mainly achieves the following three favorable merits. First, iLoc takes advantage of the fact that all data packets of the 802.11n protocol are broadcast packets, and modifies the protocol stack in the monitoring APs. This modification ensures the application layer receives the channel state information (CSI) for localization without affecting the regular communications in wireless local area network (WLAN). Second, iLoc converts the collected CSI matrixes into the Hankel matrixes and brings them into the super-resolution MUSIC algorithm to get the angle of arrival (AOA). Finally, iLoc employs triangulation for localization and correlates the MAC address and CSI of the device by modifying the Linux kernel to distinguish different mobile devices. iLoc also uses GPU, Kafka queue, and multi-threading technologies to improve the processing efficiency. Our experiments in different environments show that the localization accuracy of iLoc for mobile devices can reach 90cm on average.