A taxonomy of service discovery approaches in IoT

1st Meriem Achir 2st Abdelkrim Abdelli 3nd Lynda Mokdad

Abstract—Nowadays, Internet of things (IoT) is becoming the the majority of existing surveys providing taxonomies are
trademark technology for connecting real-world physical entities focusing only on specific elements such as: context, data, etc.
to provide specific services and enable the creation of a smart In this paper, we aim at providing a more exhaustive and
world. Due to the growing number of connected objects and the
lack of standardization that meet IoT requirements, discovering elaborated taxonomy that includes several aspects and criteria
and selecting the desirable service is becoming a challenging to classify service-object discovery and selection approaches
issue. In the literature, many works have been put forward in the IoT-cloud domain. We provide a full comparison of the
to address service discovery in IoT. In this paper, we propose most recent IoT service discovery and selection approaches
a complete classification of these works to provide a more based on the proposed classification scheme.
elaborated taxonomy that includes several aspects and criteria.
Then, we compare some of the most recent approaches according To the best of our knowledge, few works have been pro-
to several performance criteria, followed with a discussion to set posed in the literature to survey SD approaches in the Cloud-
up the future research directions. IoT domain. Among them, only few have provided a general
Index Terms—Taxonomy, Service Discovery, IoT, Classifica- classification and a comprehensive overview of the existing
tion, Service Selection, Comparison. techniques to understand their nature and to facilitate their
description [2] [3] [4] [5] [6].
I. I NTRODUCTION
Zorgati et al [2] have classified IoT SD techniques into two
Internet of Things (IoT) is transforming our daily lives main categories: those that are protocol-based (CoAP, MQTT
through a variety of applications that cover several domains, and mDNS/DNS-SD) which offer syntactic resource discovery,
including: smart homes, smart cars, smart buildings, etc. and those that are semantic-based. They compared a number
Concretely, IoT is a paradigm in which a huge number of real- of existing works based on six criteria. However, this work
world physical entities can be connected to Internet to provide doesn’t cover all important protocols used in IoT SD and
specific services. These entities generate a large amount of doesn’t detail service description methods of each solution.
data and real time interactions, that solicit hugely Internet in- In [3], Roopa et al presented the fundamental concepts
frastructures capacities. Therefore, the Cloud Computing (CC) of SIoT paradigm (Social IoT), the prerequisites, the main
paradigm has been considered and re-designed to improve the challenges to overcome, as well as different use cases based
management of IoT systems. Indeed, the integration of the CC on these concepts. They introduce three taxonomies for SIoT
in the IoT offers a great scalability and flexibility regarding the according to: SD mechanisms, Network link selection, and
increasing number of connected objects. Furthermore, the data Trust composition. Although this survey is interesting as it
generated and collected from these objects can be transmitted introduces new criteria, it focuses only on social objects and
to the cloud for processing and analysis [1]. does not investigate other existing aspects. In [4], Abdellatif
With the advent and the development of different paradigms et al defined a new classification scheme based on different
and technologies associated with IoT, and CC, we observed a criteria: the type of architecture (centralized, distributed or hi-
proliferation of a large number of Internet-connected objects, erarchical), the operating mode (push/pull), the discovery tool
offering a wide range of services with different functionalities interface (Search engine, Recommendation systems, Global
and performances. In addition, the lack of standardization in directory), the discovery scope (remote and/or local) and the
terms of architectures and new technologies that meet IoT dependency to Internet protocols. Other essential aspects of
requirements, has become an important challenge to overcome. Cloud-IoT SD and selection such as service description and
More particularly, there is a need in terms of designing discovery mechanisms are not included in this review. In [5],
efficient techniques for the discovery and the selection of the authors classify SD techniques into two main categories:
services and objects while dealing with IoT constraints, as data-oriented and object-oriented. Data oriented solutions can
well as those of the users in an optimal way by improving the be classified into two types : Content-based, which focus
QoS (Quality of Service) and the QoE (Quality of Experience). on observational and measurement data read by sensors, and
To identify the lacks and fix some the issues, a clear and context-based, which focus on data describing the situation and
exhaustive classification of IoT service discovery (SD) and state of IoT objects. Object-oriented solutions can be classified
selection approaches is more than necessary. In the literature, into three types: location-based, geographic or relative to the
978-1-7281-9015-0/20/$31.00 © 2020 IEEE IoT objects in the network, user-oriented, which takes into
 Fig. 1. The proposed taxonomy of IoT SD approaches.

account the user’s preferences and social relationships with the the subjective aspect by introducing two new categories:
objects, and finally heterogeneous, which uses several types User requirements based and User feedback based plus the
of data representation, status description and other attributes Social-based category. And finally, regarding the protocol-
of the IoT objects. Aziez et al [6] proposed a multi-criteria based class, we consider the protocols most commonly used
classification of Cloud-IoT services according to five different in the IoT SD and classify them into three categories: service-
aspects: level of complexity, service capability, physical loca- oriented (DPWS, DNS), message-oriented (MQTT, XMPP)
tion of the SD, architectural aspect, and service representation. and resource-oriented (REST, CoAP). So, our taxonomy gives
A taxonomy is elaborated to classify the different approaches a different hierarchy’s levels compared to the existing ones
according to four perspectives: description method, discovery and it’s more extended. In the sequel, we describe each class
mechanism, architecture and context-awareness (supported or and discuss its features.
non supported). However, some used concepts and paradigms
are confused in this classification. A. Semantic-based approaches
The remainder of the paper is organized as follows. Section This class uses semantic Web techniques to discover IoT
II is devoted to present our taxonomy scheme. The following services such as: semantic similarity and relativity, ontologies
section provides a full comparison of the most recent Cloud- etc. Only few works can be mentioned within this class. Jia
IoT SD approaches. The last section concludes this paper and et al. proposed a centralized four-layer model for IoT SD via
highlights the future research directions and challenges. a multi-stage semantic service matching [7]. Also, they put
forward a hybrid service matching degree measurement by
II. T HE PROPOSED TAXONOMY calculating the concept logic and the semantic similarity for
SD mechanisms are fundamental to locate the most appro- each layer separately. In [8], Zhao et al. proposed a multi-
priate services that meet user requirements while taking into dimensional semantic-based model to describe IoT services
account the dynamic nature of the IoT environment. In this and measure the similarity between them. Besides, a density-
section, we propose a classification of the existing approaches peaks-based clustering is employed to gather similar services
through the taxonomy depicted in Fig.1. according to similarity measurement.
As we can see, we classify the existing techniques into five
classes according to the main paradigm used in the discovery B. Bio-inspired based approaches
process: Semantic based, bio-inspired based, protocol based, The bio-inspired computing paradigms have emerged be-
context based and QoS based. Then, each of these classes is cause of their inherent ability to operate without central
divided into several sub-categories. First of all, we consider control, their efficient management of resources, and their self-
the most frequent QoS parameters used in the IoT domain organization. Some works have considered the use of these
to introduce the QoS-based sub-categories, which are: Secu- paradigms in IoT SD: Rapti et al. proposed a solution based on
rity based, energy consumption based, network-infrastructure Artificial Potential Fields (APFs) [9], in which, each device in
based, as well as a hybrid sub-category containing the so- the network represents a Service Provision Nodes (SPN) that
lutions where several QoS parameters are used. Secondly, provides its functionality through services that are handled
concerning the context-based class, we take into account internally by software agents. In [10], Bao et al. developed
the objective context aspect by focusing on two categories USPIOT (Ubiquitous Services Platform for IoT) which adopts
of approaches: Location-based and Event-based, as well as an XML-based service description method and provides a
selection algorithm to find the most appropriate service quickly messages are controlled by centralized entities called, brokers.
and accurately based on Artificial Neural Network (ANN) and Among these protocols we can quote MQTT and XMPP.
three types of context (user, service and network contexts). a- MQTT-based approaches: Message Queuing Telemetry
Transport (MQTT) is a lightweight publish/subscribe messag-
C. Protocol-based approaches ing protocol designed for constrained devices and networks
Different protocols offering SD capabilities have been pro- with low bandwidth, high latency or low reliability. In MQTT,
posed as IoT standards by different organizations such as the several clients connect to a single server (broker) to publish
Object Management Group (OMG), the Internet Engineering information or subscribe to already published services. In
Task Force (IETF) etc. According to [11], there are three main [16], Kim et al. presented a MQTT-based context-aware au-
approaches for protocols used in IoT environments: Service- tonomous SD in oneM2M (Machine to machine) architecture
oriented, message-oriented and resource-oriented. to provide an auto-configuration mechanism by implement-
1) Service-oriented approaches (SOA): In SOA, a service ing an asynchronous request/response scheme using MQTT
is a discrete unit of functionality that can be accessed and publish/subscribe messaging pattern. In the proposed system,
updated remotely and its description is stored in a directory devices register the service in the resource directory based
on a server. Within this class we can report two clusters of on discovering the gateway using DNS-SD/mDNS. The gate-
approaches, DPWS and DNS based approaches. way autonomously interacts with devices using semantic Web
a- DPWS-based approaches: The Devices Profile for Web technology. After the registration step, the devices become
Services (DPWS) is a subset of web services (WS) standards oneM2M resources and are able to access devices and discover
that allow minimal interaction with WS running on the device. services using MQTT. In [17], Pareira et al. developed a
It allows devices to be dynamically found when they connect distributed MQTT-based resource discovery architecture for
to the network, and to dynamically retrieve metadata about the M2M communication (MQTT-RD), which allows the zero-
device and its hosted services. In [12], Guinard et al. imple- configuration networking in the IoT environment by introduc-
mented a process, called Real-World SD and a Provisioning ing plug and play capabilities to devices.
Process (RSDPP) to assist the developers at design time in SD. b- XMPP-based approaches: Extensible Messaging and
Besides, DPWS techniques are used for dynamically register Presence Protocol (XMPP) is an XML-based communication
and discover devices and the services they provide through a protocol for lightweight message-oriented middleware.Several
RESTful API. In [13], Son Han et al. extended the DPWS XMPP extensions are part of the experimental implemen-
standard by using a REST proxy to overcome its limits and to tations in IoT: Efficient XML Interchange (EXI) Format,
reduce the latency as well as to simplify the global topology Sensor Data, Provisioning, Control, Concentrators, Discovery.
of using RESTful Web APIs. In [18], Wang et al. proposed a lightweight XMPP scheme that
b- DNS-based approaches: The DNS protocol represents achieves less complexity of publication and energy efficiency
a kind of distributed database containing service description for resource-constrained IoT devices by optimizing the exist-
records on a DNS server. It’s used to associate an IP address ing one to support event-driven and periodic data publications
with an Internet service. DNS-service discovery is a central- in IoT. According to the subscriber’s needs, the publisher can
ized protocol that structures DNS records to ease the SD. adjust the data information published in the server.
Service providers register services in a DNS server including 3) Resource-oriented approaches:: This class includes pro-
its description A complementary protocol to DNS-SD, is the tocols based on REpresentational State Transfer (REST) archi-
Multicast DNS-SD. mDNS-SD is a distributed version of DNS tecture. It is divided in REST-based and CoAP-based.
protocol that uses the same service description structure as a- REST-based approaches: In REST, any resource can be
DNS-SD and it proactively discovers, advertises and unadver- addressed and accessed with a unique standard identifier via
tises new services that become available in the environment. URLs using HTTP protocol. In [19], Khodadadi et al. devel-
This protocol requires a minimal configuration, and operates oped a framework to define, discover and compose things and
in infrastructure-less environments while being resilient to their services which are exposed using RESTful API Modeling
network failures. Some interesting works can be quoted within Language (RAML). The discovery process is implemented as
this class: In [14], Mahyoub et al. proposed an optimized, a RESTful WS. A flow composition was considered to achieve
distributed and autonomous discovery mechanisms for smart a complex functionality by chaining loT service calls together
objects to minimize the overhead caused by some operations to combine and orchestrate the obtained flows.
of mDNS/DNS-SD protocols to improve the network traffic b- CoAP-based approaches : Constrained Application
and power consumption. In another work [15], Stolikj et al. Protocol (CoAP) is a REST-based web transfer protocol which
presented an extension of the mDNS/DNS-SD protocol with adopts request/response model using HTTP methods. The
a context-based model for describing and discovering services client can send requests that are processed by the CoAP
that enables better scalability in large networks. server. Service descriptions contain a URI, a service type, and
2) Message-oriented approaches: This class of approaches optionally a list of client-defined key-value attributes. In [20],
considers the provision and the transfer of data and asyn- Vandana et al. presented a semantic-based CoAP enrichment
chronous messages between the distributed devices to dis- for Resource Discovery (S-CoAP) based on enhancing CORE
cover services. The reliability and the QoS of the exchanged Link Format attributes so that the user request and resource
description can be handled semantically. Moreover, Ferdousi to autonomously build their social relationships to discover
et al. designed a cloud-based middleware for CoAP-based IoT other ”friendly” objects and their services. Pruthvi et al.
SD [21], which is dynamically responsive for continuously implemented an SIoT platform for smart college environment
changing constrained IoT network while providing mobility where objects are able to create their own relationships based
and interoperability. In [22], Tanganelli et al. presented an on rules set by their owners, thus minimizing human inter-
edge-centric distributed architecture to provide IoT resource vention, and to create communities [28]. This platform builds
and SD based on a distributed hash table maintained by a social profiles for objects based on the received data from
IoT gateways deployed in fog nodes which leverage standard IoT application. In [29], Xia et al. developed a decentralized
solutions including the CoRE RD and the CoAP protocol. social-like semantic-aware SD mechanism for large-scale IoT,
namely SLSA, to design a social network among devices based
D. Context-based approaches on restricted contact graph allowing these devices to interact
This class allows to discover services based on the context in an autonomous manner. OWL method is used to describe
that can be categorised into two subclasses: Objective context services, which are ranked by calculating their correlation
which refers to SD techniques that are centred on the physical degree using the fuzzy logic method. The SLSA implements
environment surrounding the user; and the Subjective context an adaptive forwarding strategy, where the service query is
which regroups approaches that are mainly oriented on users forwarded to a selected subset of neighboring devices in a
and their social relationships with IoT objects. preferred order. Kowshalya et al. proposed in [30] a com-
1) Objective Context: . munity detection algorithm based on preferences, movement
a- Location-based approaches: The service is selected and social similarity, and a flexible service search within
or recommended based on the geographical coordinates or the detected SIoT communities through intra/inter-community
relative references of the user. In [23], Wang et al. established algorithms. In [31], Yuan et al. proposed a self-organized de-
a unified IoT SD framework based on a modified version of centralized social network featured by P2P infrastructure with
Auto-ID Center’s Object Name Service (ONS) architecture a swarm intelligence search strategy for IoT SD. Moreover, a
and the Physical Markup Language (PML) with the ability to homophily-based user model is introduced to select promising
provide a notice-and-consent for IoT. Moreover, they present neighbors in each routing step considering social relationships
a mobile application interacting with this framework to notify as well as content semantic features.
users of interesting nearby services, or about existing running b- User-requirements based approaches: This class is
services that are collecting their data secretly. In [24], Tao based on user requirements to provide the most appropriate
et al. proposed a location-based trustworthy services rec- service. Pattar et al. [32] presented a progressive SD mech-
ommendation in cooperative-communication-enabled Internet anism and two strategies that use semantic and proximity
of Vehicles (IoV) to satisfy user’s personalized and precise measures to match user’s requirements with IoT resources
service requirements considering the space-time characteristics attributes. In [33], Butt et al. implemented an adaptive and
and direct and indirect friendships among users. In this work, context-aware (user- and network-based context) SD protocol
a proprietary cooperative communication architecture and a for IoT, called Trendy. This protocol employs CoAP for de-
protocol stack are designed for the IoV. scribing services, a demand-based adaptive timer and caching
b- Event-based approaches: In this class, SD is based on mechanism to reduce the communication overhead and energy
event happening in the real-world. Yachir et al. [25] proposed consumption and a context-based grouping technique based on
a service-oriented, user-centered and event-aware framework location to offer service composition and improve scalability.
for ambient services and events monitoring in ambient envi- c- User-feedback based approaches: The approaches are
ronments. This monitoring supports self-adaptation to unpre- based on user experience, comments and feedback to select
dictable changes using services replacement and replanning in the appropriate service. Quan et al. [34] proposed an IoT
case of failure. In [26], Gupta et al. implemented an event- service selection model by aggregating information from both
driven SOA-based architecture (EDSOA) for IoT applications the latest feedback from users and service objective perfor-
which is an extension of broker-based architecture to support mance analysis. This approach used a reinforcement learning
real-time, event-driven, and active service execution. In [27], algorithm to learn all subjective and objective assessments
Cheng et al. proposed an event-driven and service-oriented and consider the similarity between users. Cabrera et al.
situation-aware IoT service coordination platform, a situa- proposed in [35] a service composition model with minimal
tional event definition language, an automaton-based situa- human intervention using consumers’ feedback to determine if
tional event detection algorithm and a situational event-driven a discovered plan was correct or incorrect in order to improve
service coordination behavior model. A reliable real-time data the research accuracy and reduce latency.
distribution model is implemented based on publish/subscribe
brokers for efficient topic searching. E. QoS-based approaches
2) Subjective Context: . In this class, services are discovered to users according to
a- Social-based approaches: This class adopts a new their QoS requirements (non-functional requirements of IoT)
paradigm in IoT called Social Internet of Things (SIoT) which using specific parameters. A great amount of research and
uses social networks aspects for the IoT and allows objects effort have been investigated to enhance the QoS in IoT SD.
In our taxonomy, we consider four subclasses for this class: 4) Hybrid approaches: Li et al. developed in [42] a decen-
Security, energy consumption, network infrastructure, and a tralized Trustworthy context and QoS-aware SD framework
hybrid subclass that covers approaches using multiple QoS using an ontological model to present and match service with
parameters. context and QoS requirements. A trust propagation mechanism
1) Security-based approaches: This class is based on secu- collecting feedback from services consumers was proposed
rity mechanisms to discover services, such as authentication, to help users to select trustworthy services without adding
trust, privacy and object access control. Lee et al. implemented additional infrastructure overhead.
in [36] a web-based system allowing users to easily browse III. D ISCUSSION
nearby IoT services, understand their privacy implications
by introducing an information architecture describing their In this section, we compare some of the presented so-
privacy properties and the personal information that may be lutions for IoT SD. For this purpose, We consider several
inferred. It allows also to inspect the privacy properties by criteria: Service description method; discovery mechanisms;
controlling the collection and usage of sensor data according to architecture and IoT requirements. Service description method
the user’s privacy preferences and collaboratively evaluating its represents the way an IoT service is described so that it
potential privacy risks. Dahbi et al. proposed in [37] a secured is machine-interpretable. We can distinguish two main cat-
and distributed SD architecture for supply chain domain in IoT. egories: syntactic (JSON, key-value, XML...) and semantic
This approach is based on DHT techniques and P2P systems (OWL-S, ontologies...). It is clear that semantic methods allow
to solve some issues observed in ONS, with a focus on the a more precise and accurate discovery compared to syntactic
EPCglobal Network. Two security schemes were proposed methods. However, the use of semantic methods in a dynamic
to detect suspicious lookup queries one of them is based and a large scale system such as IoT can lead to additional
on statistical model inspired from Gaussian model and the computational costs and a high volume of network traffic. The
other is a probabilistic model based on Continuous Hidden architectural aspect can refer to the architectural design (cen-
Markov Model (CHMM). The security requirements taking tralized, distributed or hybrid) or to the implementation style
into account in this work are: authentication, data integrity (layered, agent and P2P overlay). Centralized solutions are
and confidentiality. based on a central middleware/server containing the resource
description registers, and therefore a single point of failure,
2) Energy consumption based approaches: Khanouche et fault-tolerant with security risks. For this purpose, distributed
al. proposed in [38] an energy-centered and QoS-aware ser- and hybrid solutions have been proposed to bring more flex-
vices selection algorithm for IoT services composition which ibility and resilience to IoT environments. In addition, P2P
aims at maximizing energy conservation to ensure a high networks allow a distributed, robust and scalable SD.
services availability. This approach is carried out in two SD mechanisms aim to find the desirable services for users
main phases : 1) services preselection offering the QoS level while respecting IoT requirements. After reviewing the works
required using a lexicographic optimization strategy and QoS presented in the previous section, we noticed that QoS-based
constraints relaxation technique; and 2) the selection of the solutions considerably improve the quality and reliability
most suitable services among the preselected ones for compo- of the required services delivery. Thus, to meet some IoT
sition using the concept of relative dominance of services in requirements, they consider a number of QoS parameters.
the sense of Pareto. The relative dominance of a candidate Among these solutions we quote: the approaches [38] and
service depends on its energy profile, QoS attributes and [39] conserve and minimize the rate of energy consumption
user’s preferences. Sharma et al. presented in [39] an energy which improves the availability of battery-constrained objects,
efficient architecture for device discovery in 5G-based IoT the solutions [40] and [41] aim to optimize the network
and Body sensor networks using multiple Unmanned Aerial infrastructure (throughput, latency, memory, network traffic)
Vehicles. The proposed mechanism utilizes XML charts to which increases network performance and data routing and
perform device discovery on the basis of networks state cost minimizes overhead, and finally the security-based approaches
and available energy. [36] and [37] try to ensure privacy, trust, authentication,
3) Network-infrastructure based approaches: Maiti et al. data integrity, etc.To improve trust-management and network’s
proposed in [40] an efficient approach of latency-aware fog navigability, the SIoT paradigm was introduced in [28], [29],
smart gateways deployment for IoT services which optimizes [30] and [31] by allowing spontaneously create trusted IoT
the number of fog nodes to reduce the latency induced by object communities based on context collection, rules set
traffic aggregation and processing. Antonini et al. proposed by owners, object profile similarity, etc. In order to ensure
in [41] a lightweight multicast forwarding mechanism for user satisfaction and improve QoE, User feedback and User
SD in Low-power and Lossy IoT Networks (LLN) with requirements solutions were introduced to reduce the search
low memory footprint and zero-configuration operations. The space and hence the response time. Location and event based
proposed protocol relies on filtered local flooding, which solutions aim to improve dynamicity and context-awareness.
adapts well to duty-cycled devices operating in LLNs with To provide an autonomous discovery systems, DNS-based
RPL and introduces Bloom filters to detect duplicate packets solutions was proposed. However, the main drawback of these
and prevent forward loops. solutions is the need for caching DNS entries especially for
resource-constrained devices. This issues can be solved by [16] G. Kim, S. Kang, J. Park & K. Chung, ”A MQTT-Based Context-Aware
timing the cache for a specific interval [1]. According to Autonomous System in oneM2M Architecture”, IEEE Internet of Things
Journal, 2019.
[43], the solutions that are based on CoAP, REST and MQTT [17] E. Pereira, R. Pinto, J. Reis & G. Gonçalves, ” MQTT-RD : A MQTT
protocols achieve a better interoperability requirement as well based Resource Discovery for Machine to Machine Communication”,
as the semantic solutions. However, regarding MQTT-based In IoTBDS 2019, pages 115-124.
[18] H. Wang, Member, D. Xiong, P. Wang, & Y. Liu, ”Lightweight XMPP
solutions, the MQTT clients should have an open connection Publish/Subscribe Scheme for Resource-Constrained IoT Devices”, 2017
to the broker all the time, so the MQTT broker becomes a IEEE Access, 5, 16393–16405.
single point of failure. Besides, the semantic approaches suffer [19] F. Khodadadi, A. V. Dastjerdi & R. Buyya, ”Simurgh : A Framework for
Effective Discovery, Programming, and Integration of Services Exposed
from performance issues due to the huge amount of data. in loT”, RIOT Singapore, 7-9 April 2015.
Also, XMPP protocol has a considerable overhead due to the [20] C. P. Vandana & A. A. Chikkamannur, ”S-COAP: Semantic Enrichment
streaming of XML. Besides, DPWS protocol is suitable for of COAP for Resource Discovery”, SN Computer Science (2020).
[21] R. Ferdousi & P. K. Mandal, ”LOAMY : a Cloud-based Middleware for
resource-constrained devices and improves dynamicity. CoAP-based IoT Service Discovery”, 2019 ICACCP.
[22] G. Tanganelli, C. Vallati & E. Mingozzi, “Edge-Centric Distributed
IV. C ONCLUSION Discovery and Access in the Internet of Things”, IEEE Internet of Things
Journal, vol. 5, no. 1, pp. 425438, 2018.
In this paper, we have proposed a new taxonomy for SD [23] E. Wang & and R. Chow,”What Can I Do Here? IoT Service Discovery
in IoT that classify the existing approaches according to the in Smart Cities”, IEEE CASCITS, 2016.
main paradigms used in the SD process. Then, we compared [24] M. Tao, W. Wei, S. Huang,”Location-based trustworthy services recom-
mendation in cooperative-communication-enabled Internet of Vehicles”,
the most recent and important works considering four per- Network and Computer Applications 126 (2019) 1–11.
spectives: Service description method, Discovery mechanism, [25] A. Yachir, Y. Amirat, A. Chibani & N. Badache,”Event-Aware Frame-
architecture and some IoT requirements. Also, we focused on work for Dynamic Services Discovery and Selection in the Context of
Ambient Intelligence and Internet of Things”, IEEE TRANSACTIONS
the advantages and drawbacks of each class presented in the ON AUTOMATION SCIENCE AND ENGINEERING, VOL.13, NO.1,
taxonomy to highlight some challenges and future research JAN 2016.
directives. [26] P. Gupta, T. P. Mokal, D.D. Shah & K.V.V. Satyanarayana, ”Event-
Driven SOA-Based IoT Architecture”, ICICA, vol 632, Singapore, 2018.
[27] B. Cheng, D. Zhu, S. Zhao & J. Chen, ”Situation-Aware IoT Service Co-
R EFERENCES ordination Using the Event-Driven SOA Paradigm”, IEEE TNSM,2016.
[1] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and M. [28] M. Pruthvi, S. Karthika N. Bhalaji, ”A Novel Framework for SIoT
Ayyash, ”Internet of Things : A Survey on Enabling Technologies, College”, ICCIDS, 2019.
Protocols, and Applications”, IEEE COMMUNICATION SURVEYS & [29] H. Xia, et al. ”An efficient social-like semantic-aware service discovery
TUTORIALS, VOL. 17, NO. 4, FOURTH QUARTER 2015. mechanism for large-scale Internet of Things”, Computer Networks 152
[2] H. Zorgati, R. Ben Djemaa, and I. A. Ben Amor, ”Service discovery (2019) 210–220.
techniques in Internet of Things : a survey”, IEEE SMC 2019. [30] A.M. Kowshalya, X.Z Gao and Valarmathi ML, ”Efficient service search
[3] M.S. Roopa, et al. ”Social Internet of Things (SIoT): Foundations, thrust among Social Internet of Things through construction of communities”,
areas, systematic review and future directions”, Computer Communica- Cyber-Physical Systems (2019).
tions 139 (2019) 32–57, 2019. [31] B. Yuan, L. Liu & N. Antonopoulos, ”Efficient service discovery in
[4] Abdellatif, S., Tibermacine, O. and Bachir, A. ”Service discovery in the decentralized online social networks”, Future Generation Computer
internet of things : a survey”, ISMICS,pp. 66-74 (2018). Systems 86 775–791,2017.
[5] S. Pattar, et al. ”Searching for the IoT Resources : Fundamentals, [32] S. Pattar, et al. ”Progressive Search Algorithm for Service Discovery in
Requirements, Comprehensive Review and Future Directions”, IEEE an IoT Ecosystem”, GreenCom 2019.
Communications Surveys & Tutorials, 2018. [33] T.A.Butt et al. ”Adaptive and Context-Aware Service Discovery for the
[6] M. Aziez, S. Benharzallah, H. Bennoui, ”A full comparison study of ser- Internet of Things”, LNCS, 36–47 2013.
vice discovery approaches for internet of things”, Pervasive Computing [34] H. Quan, R. Takahashi and F. Yoshiaki, ”Dynamic Service Selection
and Communications Vol. 15 No. 1, 2019 pp. 30-56. based on User Feedback in the IoT Environment”, CITS 2019.
[7] B. Jia, W. Li & T. Zhou, ”A Centralized Service Discovery Algorithm [35] C.Cabrera, A.Palade, G.White & S.Clarke ”Services in IoT: A Service
Via Multi-stage Semantic Service Matching in Internet of Things”, 2017 Planning Model Based on Consumer Feedback” ICSOC, 304-313 2018.
IEEE CSE. [36] H.Lee, et al. ”IoT Service Store: A Web-based System for Privacy-aware
[8] S. Zhao, L. Yu, B. Cheng & J. Chen, ”IoT Service Clustering for IoT Service Discovery and Interaction”, CoSDEO’18.
Dynamic Service Matchmaking”, MDPI Sensors 2017, 17, 1727. [37] A. Dahbi, ”Supply Chain Discovery Services in an Internet of Things
[9] E. Rapti, A. Karageorgos, C. Houstis & E. Houstis, ”Decentralized ser- Environment”, Thesis Univ Ottawa Canada, 2017.
vice discovery and selection in Internet of Things applications based on [38] M. E. Khanouche, et al. ”Energy-Centered and QoS-Aware Services Se-
artificial potential fields”, Service Oriented Computing and Applications, lection for Internet of Things”, 2016 IEEE TRANS ON AUTOMATION
11(1), 75–86, (2016). SCIENCE AND ENGINEERING.
[10] J. Bao, Y. Ding & H. Hu, ”A New Service Selection Algorithm in [39] V. Sharma, F. Song, I. Youa & M. Atiquzzaman, ”Energy efficient device
USPIOT”, IEEE CSAE, 2012.pp.22-26. discovery for reliable communication in 5G-based IoT and BSNs using
[11] K. Fysarakis, et al. ”Which IoT Protocol ? Comparing standardized unmanned aerial vehicles”, JNCA 97 (2017) 79–95.
approaches over a common M2M application”, 2016 IEEE GLOBCOM. [40] P. Maiti, H. K. Apat, B. Sahoo & A. K. Turuk,”An effective approach
[12] D. Guinard, et al. ”Interacting with the SOA-Based Internet of Things of latency-aware fog smart gateways deployment for IoT services ”,
: Discovery, Query, Selection, and On-Demand Provisioning of Web Internet of Things 8 (2019) 10 0 091.
Services”, IEEE TRANSACTIONS ON SERVICES COMPUTING, [41] M. Antonini, et al. ”Lightweight multicast forwarding for service dis-
VOL.3, NO.3, 2010. covery in low-power IoT networks”, SoftCOM 2014.
[13] N. Son Han, S. Park, G. M. Lee & No. Crespi, ”Extending the Devices [42] J.LI, Y.BAI, N.ZAMAN and V.C.M.LEUNG, ”A Decentralized Trust-
Profile for Web Services (DPWS) standard using a REST proxy”, IEEE worthy Context and QoS-Aware Service Discovery Framework for the
Internet Computing, 2015, 19 (1), pp.10 - 17. Internet of Things”, INTELLIGENT SYSTEMS FOR THE INTERNET
[14] M. Mahyoub, A. Mahmoud & T. Sheltami, ”an optimized discovery OF THINGS, 2017.
mechanism for smart objects in IoT”, IEEE IEMCON, 2017. [43] Y. Fathy, P. Barnaghi & R. Tafazolli, ”Large-scale indexing, discovery
[15] M. Stolikj, et al. ”Context Based Service Discovery in Unmanaged and ranking for the Internet of Things (IoT)”, ACM Computing Surveys,
Networks Using mDNS/DNS-SD”, 2016 IEEE ICCE. vol. 51, no. 2, pp. 1–67, 2018.