Contextual decision making and action enforcing applications in wireless networks and IoT using SDN as a platformKarimi, Maryam (2022) Contextual decision making and action enforcing applications in wireless networks and IoT using SDN as a platform. Doctoral Dissertation, University of Pittsburgh. (Unpublished)
AbstractWith the rise of the Internet of Things (IoT), new challenges and opportunities have emerged, providing the potential to improve living standards, higher efficiencies and lower costs. A variety of wearable devices and sophisticated, yet not very expensive connected devices are now able to gather different kinds of information (mostly) on their premises and enable people to make decisions or control actions. The complexity of such decisions need to be balanced with performance tradeoffs. Different applications may need different levels of service that require different communication, storage, and service costs. A suitable architecture that can employ data driven decisions for performance improvements comprises of software defined networks (SDN) and software defined perimeter (SDP). SDN and SDP are suitable platforms due to their ability to have a view of the network traffic for implementing the services that require gathering information from the context and performing contextual decision making while balancing performance tradeoffs, managing specific parameters, and enforcing actions based on them. This dissertation examines three case studies that look at such performance tradeoffs. In the first case study, we explore the use of historical data in WiFi networks to create a classification QoS decision tree that predicts the maximum delay due to specific traffic situations with specific context parameters and makes rapid decisions possible to manage wireless network resources considering quality requirements. We use OpenFlow network access and gathering necessary context in wireless networks. The tradeoffs here involve balancing traffic flows from WiFi access points to meet latency constraints of end devices. In the second case study, we investigate contextual integrity verification in IoT where we look at ``levels" of integrity verification. A variety of IoT devices may be required to outsource sensed or generated data to multiple heterogeneous cloud servers. We posit that it is the Data Owner’s responsibility to verify whether the stored data remain unchanged. However, the “level” of this verification may be different under different contexts. We propose four typically disparate methods of integrity verification and consider the “toll” in terms of time, storage and communication to decide on a suitable data integrity verification process. We adapt the notion of contextual privacy to extract important parameters that determine the right level of integrity to be applied to data blocks. Such contextual information can be extracted in a SDN while security context information and a secure infrastructure for authentication and communication is possible through a secure architecture with the integration of SDP and SDN. Finally, in the third study, we investigate "levels" of reliability of contextual data along with the use of Bayesian Regression to enhance decision making and determine how critical each context variable is and how it would affect choosing appropriate parameters for action to perform. Again, SDN and SDP can provide the contextual information needed to maintain the various reliability levels. Share
Details
MetricsMonthly Views for the past 3 yearsPlum AnalyticsActions (login required)
|