|
" Service Discovery and Mobile Crowdsourcing Recruitment in Social Internet-of-Things "
Khanfor, Abdullah
Massoud, Yehia
| Document Type
|
:
|
Latin Dissertation
|
| Language of Document
|
:
|
English
|
| Record Number
|
:
|
1057591
|
| Doc. No
|
:
|
TL56708
|
| Main Entry
|
:
|
Khanfor, Abdullah
|
| Title & Author
|
:
|
Service Discovery and Mobile Crowdsourcing Recruitment in Social Internet-of-Things\ Khanfor, AbdullahMassoud, Yehia
|
| College
|
:
|
Stevens Institute of Technology
|
| Date
|
:
|
2020
|
| Degree
|
:
|
Ph.D.
|
| student score
|
:
|
2020
|
| Note
|
:
|
153 p.
|
| Abstract
|
:
|
The Internet-of-things (IoT) networks are witnessing a drastic increase over the years. Twenty billion devices connected to the Internet are anticipated in 2022. The IoT is mainly characterized as a vast network of heterogeneous connected devices that communicate with each other to handle various users' needs. Recently, the concept of Social IoT (SIoT) has put forward the idea of socialization in IoT where devices can establish social relations, similar to people's social networks, to indict a collaboration or permission between each other. Despite having these relations that benefit in raising trustworthiness among SIoT devices, conceiving efficient service discovery procedures in such large-scale networks remains impeded by the system's increasing complexity. Therefore, there is a pressing need to design efficient methods to downgrade the complexity of service discovery in SIoT and help exploit the devices' social relations. One of the tools we consider is graph analytic. To this end, we convert the SIoT social relations into graphs to which we apply different community detection methods, namely Louvain, Bron-Kerbosch, and Order Statistics LOcal Method (OSLOM) algorithms. The objective is to cluster the obtained graphs into several virtual communities whose members share strong social relations. Noticing that these algorithms are computationally demanding and taking into account the social relations only, we propose to implement Graph Neural Network (GNN)-based technique to embed the IoT nodes by incorporating their social relations as well as their features. Comparisons with standard methods are then performed. Afterward, we demonstrate the benefits of SIoT clustering for different smart city use-cases involving service discovery. We first investigate the case of a Natural Language Process (NLP)-based approach that handles crowdsourcing textual requests. The framework matches tags extracted from the textual requests to the suitable communities meeting the task requirement. Then, it finds the list of IoT devices accordingly. As a second use-case, we propose to leverage the community detection approaches in Spatial Mobile Crowdsourcing (SMCS), where skilled IoT users or autonomous devices are commissioned to travel toward specific locations to execute tasks physically. Given the determined communities, an Integer Linear Program (ILP) is developed and applied to this reduced search space to determine appropriate devices/workers. Finally, we execute a service discovery process using our proposed technique for mobile edge computing search. The goal is to find available IoT edge computers having sufficient capabilities to handle computational tasks. Machine learning algorithms are also employed to predict potential candidates' computational service time belonging to trustworthy SIoT communities. Through the different use-cases, we showcase low complex SIoT service discovery in enabling real-time and practical applications in smart cities. To sum up, through the different use-cases, we showcase the ability and potential of low complexity SIoT graphs that aid in service discovery that can enable real-time and practical applications that serve smart cities and large-scale IoT systems.
|
| Descriptor
|
:
|
Computer science
|
|
|
:
|
Engineering
|
|
|
:
|
Information technology
|
|
|
:
|
Systems science
|
| Added Entry
|
:
|
Massoud, Yehia
|
| Added Entry
|
:
|
Stevens Institute of Technology
|
| |