Social-based smart objects discovery and composition in internet of things

다국어 초록 (Multilingual Abstract)

Recent advances in the Internet of Things (IoT) technology have led to the rise of a new paradigm: Social Internet of Things (SIoT). The emerging Social Internet of Things has undoubtedly opened up a myriad of new business, social, and research opportunities by allowing smart objects to autonomously establish social relationships with each other and exchange information. However, the new paradigm, as inspired by the idea that smart objects will soon have a certain degree of social consciousness, is still in its infant state for several reasons.
One of the most fundamental research challenges posed by SIoT is the fact that there is yet to be a coherent social network structure for organizing and managing smart objects that elicit social-like features. In particular, a social network of smart objects must be built upon three design principles. First of all, the network structure must be designed such that it exhibits scalability in terms of embracing the explosive accumulation of smart objects and inter-object relations. Moreover, the network structure must be able to cope with the heterogeneity of smart objects as they are extremely diverse in terms of functionality and descriptions. Last but not least, the ability to efficiently adapt to the dynamicity of smart objects must be incorporated into the social network structure by reorganizing itself.
To fully understand how and to what extent these objects may mimic the behaviors of humans, there is an urgent need for a method that quantitatively estimates the link strengths of inter-object relations within the social network. In human-centric social networks, people are connected by friendship relations, which indicate the degree of their closeness and how much they share similar interests. Furthermore, these relations are augmented with trust among people so that they can be utilized for various real-world applications such as service recommendations, viral marketing and so on. Similarly, the strengths of social relationships among smart objects must be properly quantified and integrated with trust in smart object discovery and composition, so as to proliferate the provisioning of SIoT composite services.
In this thesis, we propose a hypergraph-based overlay network model for effectively organizing and managing smart objects and their social relations. This particular social network structure is a generically constructed model, which exposes various SIoT interactions at different granularity levels to ensure the flexibility and consistency of the model. We further design a smart objects discovery mechanism which utilizes the proposed SIoT network structure. Then, we introduce a model named social strength prediction model (SSPM), which infers social connections and quantitatively predicts the strengths of inter-object connections using the co-usage data of smart objects. Based on the social strength measures, we integrate the model with our trust computation algorithm, eventually yielding a trust-augmented social strength (TASS) management protocol that can support smart objects composition.
Finally, we perform various experiments to test the feasibility of the proposed network model as well as the trust-augmented social strength computation method. We built a smart home automation demo box to evaluate the social network structure in terms of the three design principles, and conducted additional experiments using real-world datasets to demonstrate the resiliency along with the accuracy of the proposed TASS protocol. Overall, our work and findings provide valuable insights to the development of future SIoT systems.