Secure Device and Knowledge Discovery in Internet of Things

With the ever-growing adoption of Internet of Things (IoT), there is continued development and deployment of new nodes with various capabilities and services. Such heterogeneity introduces complexity and a significant need for neighbor service discovery frameworks. Because the types of services available are orders of magnitude more than pre-IoT era, researchers need new techniques to accurately discover not only the topologies and adjacencies but also the capability of other nodes and neighbors. Device discovery is usually the first task to be performed, immediately after deployment, as it enables communications, scheduling, and channel estimation. The existing discovery mechanisms do not consider the introduced overhead and are designed without consideration of the limited resources and the real-time computation requirements. In this dissertation, we propose a network-aware discovery framework based on the requirements of Internet Protocol (IPv6) and Low-power Wireless Personal Area Networks (LoWPAN). We propose a resource-oriented discovery protocol to achieve high accuracy while reducing coordination overhead and communication overhead imposed by task division or centralization of data mining with uncertain data provenance. This is done by utilizing computational resources available throughout various layers of data nodes and incorporating the fog computing paradigm, thus pushing decisions to the network edge and endpoint devices. Particularly, this dissertation looks at the problem of neighbor device and service discovery at Layer 3 and Layer 4. To date, no systematic investigation has considered the transport layer header for neighbor service discovery. We employed multiple methods to understand the challenges encountered by IoT designs and tested the proposed framework using software simulations and high-level simulation of IoT nodes and motes. Based on the experiments, it is demonstrated that network and transport layer headers are suitable for discovering devices and services in IoT networks with only a small number of overhead messages without much impact on the convergence time or the end-to-end delay and jitter. The findings provide support for the argument that device discovery can be enhanced using a cross-layer approach