Abstract
The Internet of Things (IoT) has developed rapidly with the aim to improve the quality of modern life. Despite its considerable benefits, the IoT system poses many design challenges, among which assessing an IoT system's reliability is critical for assuring the success rate of IoT service delivery. As a cyber-physical system consisting of physical and communication subsystems, an IoT system exhibits various dependent and dynamic behaviors that significantly complicate its reliability modeling and analysis. Particularly, functional dependence (FDEP) exists in IoT systems, where the failure of one component (trigger) causes other components (dependent components) to become isolated (inaccessible or unusable). When such an isolation effect occurs with a certain probability, it is referred to as the probabilistic functional dependence (PDEP) behavior. The FDEP or PDEP behavior can cause deterministic or probabilistic competitions in the time domain between failure isolation and failure propagation effects, making reliability analysis of IoT systems challenging. Existing works addressing the competing failure effects have assumed that any failure propagation originating from a component instantaneously takes effect, which is often not true in real-world scenarios. Moreover, the existing works have assumed non-cascading FDEP or PDEP, where each system component can be a trigger or a dependent component, but not both. However, in practical systems with hierarchical configurations, cascading FDEP or PDEP can take place where a component plays a dual role as both a trigger and a dependent component simultaneously. Such a component causes correlations among different FDEP or PDEP groups, further complicating the reliability analysis of IoT systems. This dissertation makes contributions by proposing combinatorial methodologies for evaluating the reliability of IoT systems subject to cascading competing failures and random propagation time. This dissertation also contributes by addressing realistic random isolation time for IoT systems subject to the FDEP behavior. The suggested methodologies are flexible without limitation on distribution types of component lifetime, failure propagation time, or isolation time/factors. Detailed case studies on smart home systems and body sensor networks are conducted to illustrate the application and advantages of the proposed approaches. Effects of different model parameters on the reliability of IoT systems are investigated. The correctness of the proposed methods is verified using the continuous-time Markov chain-based method.