An efficient social-like semantic-aware service discovery mechanism for large-scale Internet of Things

Abstract

Due to the enormous search space, dynamic availability, and restrictions on geographic positions, achieving a scalable and efficient service discovery mechanism for large-scale Internet of Things (i.e., IoT) is a challenging job. Owing to the similarity between social networks and IoT, social strategies can be integrated to improve the performance of IoT solutions. In this paper, we propose an efficient social-like semantic-aware service discovery mechanism named SLSA by mimicking human-like social behaviors and exploring cooperative intelligence. Our mechanism can discover desired services in a fast and scalable manner. The update process of knowledge index adopts a dual-modular-ordering stack strategy that makes search more efficient. Considering the semantic similarity and semantic relativity of two concepts in the domain ontology, we introduce the fuzzy logic method to calculate their correlation degree for device ranking. The SLSA implements an adaptive forwarding strategy, where the service query is forwarded to a selected subset of neighboring devices in a preferred order. We conduct comprehensive experiments to evaluate four mechanisms by establishing dynamic environments. The simulation results show that the SLSA achieves better performance than the other relevant mechanisms with three aspects. Furthermore, confirmative tests are carried out on the characteristics of small-world networks.

Introduction

The paradigm of Internet of Things (i.e., IoT) connects a massive number of pervasive and heterogeneous devices in the real physical world to the Internet. With the rapid progress in the embedded system, hardware miniaturization and low-cost device manufacturing, these devices are prompted to become more and more intelligent, which are also known as smart devices [1]. An IoT application can be implemented from multiple smart devices with services. The number of smart devices deployed around us is increasing exponentially, which continuously generate and publish an enormous number of services on the Internet [2]. However, not all the services are needed in a context-aware environment. Usually, we need a specific set of services at a time that match with our queries. Service discovery is such an essential functionality that enables the network to look efficiently for an appropriate device that provides the desired service. The importance of service discovery in IoT has been emphasized in [3]. However, it is difficult to achieve such functionality, since a large number of devices results in an enormous search space, and the network traffic will become too large to be managed efficiently, giving rise to the scalability problems. Beyond these, this functionality faces other challenges, like dynamic availability, timeliness, restrictions on geographic positions, environmental context, and so on. In brief, how to achieve a fast, scalable and efficient service discovery mechanism for large-scale IoT is a challenging but fundamental job.

Recently, there have been proposed a large number of independent studies [1], [4], investigating the potentialities of integrating social networks into IoT solutions (e.g., service discovery, topology maintenance). Particularly, a new paradigm named social Internet of Things (i.e., SIoT) has been introduced in [4], which follows the same principles that characterize social networks between human beings. These smart devices in SIoT can form social connections and generate entirely new kinds of applications and services with the aid of cooperative intelligence. The inherent characteristics in social networks can be reasonably used for service discovery in IoT. An efficient service discovery mechanism should apply an effective forwarding strategy that quickly guides the queries to arrive at the target devices with minimum detours. Notably, the ‘small world’ theory refers to the existence of short chains of acquaintances among individuals in societies, making the network navigable [5]. For instance, human beings in social networks are connected by their social relationships and look for someone, directly employing some acquaintances (or friends of friends and so on) who potentially have the relevant knowledge in a distributed manner. Regarding the discovery functionality in IoT, each smart device has to gain, store, and manage the useful knowledge of its friends as human beings do, and subsequently implement the discovery operation based on the maintained valuable information. We can envision that, in the next 10–20 years, these smart devices will have the capability to discover the desired services and collaborate with each other to accomplish specific tasks autonomously.

To address the abovementioned issues, we put forward an efficient social-like semantic-aware service discovery mechanism named SLSA for large-scale IoT by mimicking human-like social behaviors and exploring cooperative intelligence. The contributions of this paper are summarized as the following six aspects:

• The SLSA is implemented in a decentralized manner, rather than a centralized mode.

• The update process of knowledge index adopts a dual-modular-ordering (i.e., DMO) stack strategy that makes search more efficient.

• The desired services are described by utilizing an ontology tree based on the OWL method. Considering both the semantic similarity and semantic relativity of two concepts between the service information contained in the query and that obtained in a specific device, we introduce the fuzzy logic method to calculate their correlation degree (i.e., CD) for device ranking.

• The SLSA implements an adaptive forwarding strategy, where the service query is forwarded to a selected subset of neighboring devices in a preferred order based on their CD rank. This strategy is adaptive according to the correlation degree of the device to discovery queries, which can utilize local network properties to maximize the benefits of global network navigation. Moreover, this strategy can alleviate both the network traffic problem and the energy consumption problem.

• Note that, it is unlike with some previous mechanisms that construct small-world behaviors based on intentionally clustering devices into different communities or groups. In contrast, the SLSA establishes a social network among devices based on restricted contact graph and allows these devices to interact with each other in an autonomous and distributed manner as human beings do. Following the guidance of this mechanism, those devices with some commonalities (e.g., the same interest) will gradually construct strong relationships with each other and form communities spontaneously.

• We conduct comprehensive experiments to evaluate four mechanisms by establishing dynamic environments. The simulation results show that the SLSA achieves better performance than the other relevant mechanisms with three aspects: success rate of queries, average number of relay nodes and average path length of searches. Furthermore, confirmative tests are carried out on the characteristics of small-world networks (i.e., average clustering coefficient and average path length of nodes).

The remainder of this paper is organized as follows: we first introduce some related work in Section 2 and explain the theoretical basis in Section 3. Then the efficient service discovery methodology is detailed described in Section 4. Simulation results and analysis are presented in Section 5. Finally, we conclude our work and highlight the future research directions in Section 6.