Research On Social Web Of Things Architecture, Model And Key Technologies Of Security And Privacy

Internet of Things (IoT) connected various sensors, actuators and embedded devices in the physical world into Internet via wired and wireless ways. IoT expended the ability of system to acquire information from the physical world, which has formed the converged abstract digital environment of human and physical world. Compared to Internet, IoT has the following characteristics:(1) deployed massive heterogeneous sensors and could acquire different data formats and contents;(2) the environmental data that IoT terminals collected has realtime performance;(3) user centric innovation is the driving force of IoT.Recent IoT platforms are tightly-coupled and the infrastructure and service ability are hardly opened and shared. Meanwhile, the resources of third parties are difficult to integrate into IoT system to expand IoT applications. Also, recent IoT platform has the disadvantages of market fragmentation, small-scale applications, long development cycle and so on, which hindered the development of IoT. Therefore, sharing IoT data and control ability anytime and anywhere, openning pervasive service ability and implementing ubiquitous computing are the future of IoT technology.According to the above problems, we concluded the future development of IoT as openness and socialization. And we proposed the Social Web of Things (SWoT) architecture in this paper. From the openness aspect, the main idea of SWoT is using the resource oriented, loosely-coupled Web REST as the architectural style, abstracting the heterogeneous devices and data into unified resource model and using Web URI to unique identify the fine-grained IoT resourcs. From the socialization aspect, the main idea of SWoT is to support intelligent interaction between human and devices. Meanwhile, we modeled the nodes and its physical and logical links in SWoT based on the Supernetwork approach. The multi-dimensional social relations make a greater possibility for resource sharing, reputation evaluation, access control and other services. Based on SWoT architecture, we developed a smart home prototype and a resource discovery application.However, large amounts of data related to people’s daily life have been collected by cameras and sensors, which could bring the privacy leakage. Especially, with the openness of IoT environment, communications and data exchanges become more frequently. Also, the third parties can acquire user data via open APIs, which threats the sensitive data. For sensor data could be stored on and managed by the distributed heterogeneous SWoT gateways, it brings the security issues. With the trend of socialization of IoT, many factors like various attributes, complex social relations and UGC would probabily unique identify a specific node. Therefore, the traditional security and privacy technics cannot be used directly in SWoT, and it is more difficult to protect privacy.According to the security and privacy issues brought by IoT socialization, we proposed the background knowledge based privacy metric algorithm. Firstly, we analyzed the background knowledge in SWoT environment. Then we designed the multivariable kernel regression model based privacy metric algorithm for the background knowledge of joint attributes and link relations. We introduced the local bandwidth and covariance matrix to advance the traditional kernel regression. According to the UGC background knowledge, we proposed linear regression and naive Bayesian classification based algorithems respectively. Considered the high dimensions of content vector, which would bring high computing costs, we adopted PCA to advance. By implementing on Weibo’s real data, it is proved that our privacy metric algorithm has good performances. According to the storage and searching security brought by the distributed SWoT gateways, we proposed a reputation based virtual security domain algorithm for security-enhanced search. We designed a converged reputation evaluation algorithm considering the underlay node’s isomerism and instability, as well as the interaction evaluation of overlay nodes. By structuring P2P Chord based virtual security domains, we discussed the security issues on resource storage and searching. Meanwhile, we discussed a two-layer access control mechanism and node selection mechanism. Based on the simulation, the algorithm we proposed has the advantages of low maintenance costs, good extensibility and could improve the security of data access and node.