The widespread deployment of wireless communication systems creates unprecedented opportunities to impact our daily lives. Regardless of whether wireless infrastructures are used just for communication or as the basis for actual responses, large-scale wireless data provide increasing opportunities for detecting environmental changes caused by moving objects. Indeed, it is expected to develop the ability to make use of existing wireless infrastructure and sensing data to track moving objects which do not carry radio devices and may not even being aware of being tracked. However, these wireless data are dynamic and have complex data characteristics, such as multi-scale, multi-source and multi-modal. As these data become large and more detailed, new challenges are emerging for intrusion learning.
This project aims to develop effective and scalable multi-modal passive intrusion learning techniques that have the capability to detect and track device-free moving objects in pervasive wireless environments through adaptive learning in a collaborative way. In contrast to traditional techniques, which require pre-deployment of specialized hardware, and thus not easily deployed for unscheduled tasks and may not be scalable, this project leads to new insights into intrusion learning by mining on wireless environmental data, as well as leading to new approaches to device-free wireless localization, which can be used to assist a broad array of applications (e.g., identification of people trapped in a fire building during emergency evacuation). Project results are expected to open a new venue for integrating learning capabilities into emerging pervasive wireless fields. The educational component seeks to equip students with the necessary background and practical skills needed to contribute to information technology and have a practical impact on a large set of cross-section domains.
The widespread deployment of wireless communication systems creates unprecedented opportunities to change the paradigm for trustworthy computing in pervasive wireless environments. Indeed, when wireless networks have been deployed with sufficient density and equipped with growing sensing capabilities, there are large-scale wireless measurement data which become available and enable the ability for wireless users to sense the world at a microscope level. A particularly promising area for enhancing trustworthy computing is to perform device-free environmental intrusion learning, i.e., detecting and tracking moving objects even when they do not carry any radio devices, by making use of pervasive wireless infrastructures and the wireless environmental measurement. Since this device-free intrusion learning ability can provide tremendous cost savings and can be enabled at any time, it is expected to have a broad array of applications, ranging from intrusion detection to emergency evacuation to battlefield and border protection. Therefore, the main focus of this project is to develop effective and scalable multi-model device-free intrusion learning techniques that are capable to characterize moving objects in mobile environments. There are two key issues in intrusion learning: detection and tracking. Detection is to determine the presence of objects and tracking is to characterize the movement patterns of moving objects. At the tracking side, we have developed algorithms for understanding movement patterns in geosensor networks. We have defined different types of movement patterns which can be exploited for understanding user activities in the networks. Specifically, we have designed the algorithms for direction-based clustering. Also, we have developed an adaptive device-free localization framework. Specifically, three speed change detection schemes were designed based on some statistical learning techniques. Also, we have provided a new passive localization algorithm, named RIG, which is based on Algebraic Reconstruction. RIG can exploit triangle-based geometric pivot points to reduce the position estimation error. At the detection side, we have conducted the research on intrusion learning from different perspectives. One goal is to provide a better understanding of human mobility patterns and develop suitable ranking metrics to evaluate the anomaly levels of activities in a given network. In these studies, we have taken both temporal and spatial nature of mobile data into the consideration. Also, we have carefully exploited various types of contextual information which can be helpful for intrusion learning. Moreover, we have studied how to detect anomaly moving activities in very noisy indoor environments. Moreover, we proposed a stochastic model for context-aware anomaly detection in indoor location traces, which were collected by the sensors attached to medical devices in a hospital environment. Along this line, we first investigated some unique properties of these location traces. We showed that they could help to capture the movement patterns of the medical devices. Finally, we showed that the evidences from multiple information sources, if properly integrated, can help to improve the detection performances. There are various educational activities. First, we have given presentations and invited talks on project-related topics at major conferences, universities, and research labs. Also, the project contributed to the training of 3 graduate students and 1 female undergraduate students, leading to employment at the University of North Carolina – Charlotte and industry positions. Moreover, some course materials of 2013 Science & Technology Summer Camp at Rutgers University are based on the research and development results of this project.