Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI
INF-NDN IoT: An Intelligent Naming and
Forwarding in Name Data Networking for
Internet of Things
GHULAM MUSA RAZA1, IHSAN ULLAH2, MUHAMMAD SALAH UD DIN3, MUHAMMAD ATIF
UR REHMAN4, and BYUNG SEO KIM, (Senior Member, IEEE)5
1,2,3,5Department of Software and Communications Engineering, Hongik University, Sejong 30016, South Korea (e-mail: ghulammusaraza96@gmail.com,
danish1852@gmail.com, slah_udin@outlook.com, jsnbs@hongik.ac.kr)
4Department of Department of Computing Mathematics, Manchester Metropolitan University, Manchester, UK(e-mail: m.atif.ur.rehman@mmu.ac.uk)
Corresponding author: Byung Seo Kim (e-mail: jsnbs@hongik.ac.kr).
This research was financially supported in part by the National Research Foundation of Korea(NRF) grant funded by the Korea
government(MSIT) (No. 2022R1A2C1003549) and in part by a 2024 Hongik University innovation support program fund.
ABSTRACT Internet of things (IoT) has emerged as a quintessential paradigm of communication
systems. Current literature introduces notion of a named data network for IoT (NDN-IoT), optimizing
IoT communication by employing name-based networking. However, the advancements introduced by this
approach are inadequate when dealing with URL-based naming and forwarding. For instance, length and
ambiguities in content names are still open challenges. In addition, the intelligent exploration of content
names to discern a forwarding clue is a significant research gap. To achieve intelligent communication,
understanding the interest name and acquiring a forwarding clue is crucial. Focusing on this gap, an
intelligent naming scheme called INF-NDN IoT is proposed that correlates with a forwarding mechanism
as well. The proposed INF-NDN IoT improves the NDN naming schemas by utilizing natural language
processing (NLP) techniques and selecting supernodes and ordinary nodes in the network. INF-NDN IoT
assigns (forwarding clue) semantic tags to content names as well as to supernodes that in turn perform the
semantic forwarding. Experimental results have shown that INF-NDN IoT outperformed existing work, and
has better results in terms of name length, name memory utilization, interest satisfaction rate, retrieval time,
hop count, and energy consumption.
INDEX TERMS Named data network, internet of things, natural language processing, supernodes, naming,
forwarding, semantic tag.
I. INTRODUCTION
Internet of Things is a vital component of today’s informa-
tion network, which has expanded the horizons of internet
services [1]. Recently, many IoT-based solutions have been
introduced in various domains [2–4], such as smart homes,
healthcare, the internet of vehicles, and smart campuses.
These applications generate huge amounts of data and require
efficient communication and computation mechanisms to
perform timely decisions. Although IP-based communica-
tion has advantages like unbounded data retrieval from IoT
devices, it also has limitations such as complex network
configuration, security, and mobility, etc.
Current literature suggests that information-centric net-
working (ICN) provides a more practical solution to IoT is-
sues like efficient communication and ensuring data security
between diverse and heterogeneous devices. Among the var-
ious ICN solutions, named data networking (NDN) demon-
strates the most potential [5] in IoT communication, owing
to its ability to dissociate content from location, in-network
caching, scalability, flexible forwarding mechanism, and
naming features. Utilizing NDN to build IoT networks has
become a popular trend in the industry [6]. NDN highlights
the availability and access to content by assigning unique
names to each piece of data. NDN content naming eases
content caching using content store (CS) implementation, ag-
gregation using pending interest table (PIT), and content for-
warding using forwarding information table (FIB). Although
NDN offers a range of possibilities that contribute to the
availability and accessibility of content, it also brings certain
technical difficulties. One of these challenges is the explo-
VOLUME 4, 2016 1
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
ration and understanding of content names. In NDN, request
routing, data retrieval, interest satisfaction rate, and content
discovery all significantly depend on names[7]. NDN name
is text data such as /korea/university/hongik.edu/computer-
science/documents/students.txt. Text resolution and length in
NDN name is the most ignored problem in NDN [8]. The ex-
isting state-of-the-art NDN-based schemes mostly utilize hi-
erarchical or URL-based naming conventions in their name-
generation process. Authors in [9] explored that memory
utilization is higher due to the URL-based names. Moreover,
FIB requires scanning tens or hundreds of characters to find
the longest-matched prefix, which is time-consuming and
resource-intensive.
Studies referenced in [10] emphasized how the text res-
olution in the NDN name is affecting the network perfor-
mance. Hierarchical naming doesn’t understand the contex-
tual meanings of names. For instance, when a word like
“apple” is used in an interest’s name, the network interprets it
merely as a literal string, disregarding its contextual meaning,
such as fruit or technology company. In addition, NDN
doesn’t remove other name ambiguities or errors such as pho-
netic similarity, spelling differences, multiple languages, ini-
tials, titles, honorifics, and out-of-order components, which
can be found in content names. All these naming errors are
even more inflated in NDN-IoT. The reason is the length
of hierarchical structures like URLs can vary due to un-
necessary characters in them. URL-based content names are
long, complex, and unrestrained formats. On the contrary,
IoT-based names are typically transient, fleeting, new, with
diverse priorities, and from various locations.
Another notable research gap in NDN is semantic for-
warding. Content names have the potential to support data
forwarding in NDN-IoT [11]. Despite this fact, vanilla NDN
follows a broadcast mechanism to forward the interest.
NDN overlooks the potential of names in providing con-
textual or semantic clues for forwarding decisions. For in-
stance, an NDN IoT scenario where a student sends an in-
terest “hongik/sports/accessories/available/resources”. Cur-
rently, the network might process this interest name as a
generic text string. However, with semantic understanding,
the network could efficiently forward the request directly
to the Hongik smart sports environment. Similarly, if a re-
quester requests interest “Prof/Kim/exams-result”. Network
after semantic awareness, could instantly forward the request
to the campus exam department’s database. Therefore, it is
important to consider name semantics in NDN to fill the
research gap of semantic forwarding. In essence, intelligent
naming can support semantic forwarding and efficient data
retrieval. By obtaining semantic information from content
names, NDN routers can forward NDN packets based on their
context, improving the efficiency and effectiveness of data
dissemination.
Faced with these challenges and research gaps, this article
proposes a unique naming and semantical forwarding scheme
for NDN IoT. First, we propose a natural language processing
(NLP) [12, 13] based naming scheme and semantic tags.
A semantic tag is a forwarding hint that is obtained from
the NDN name of interest and FIB names of supernodes.
Supernodes are a set of resource-rich nodes identified in the
network to enable semantic communication. Subsequently,
semantic tags are integrated into the namespace design of in-
terest names and assigned to supernodes.This correlation be-
tween naming and forwarding aids in establishing a cohesive
framework for semantic communication. In addition, selected
supernodes have complete knowledge of the network’s topol-
ogy and each supernode stores the same database known as
the complete database (CDB), such as [14]. The major con-
tribution of the proposed methodology is focused on naming.
To summarize, the major contributions of this work are:
1) We propose an intelligent naming and forwarding
scheme in NDN IoT, called INF-NDN IoT. INF-NDN
IoT introduces a unique naming scheme that employs
NLP to shorten the length of the content name and
effectively resolves ambiguities that may arise in the
content names.
2) INF-NDN IoT introduces supernodes to enable a se-
mantic tag-assisted forwarding strategy which leads to
semantic content retrieval. To this end, semantic infor-
mation is extracted from both the Interest/Data packet
names and the FIB names of supernodes. Subsequently,
corresponding semantic tags are assigned to both the
Interest/Data packets and the supernodes. As a result,
packets are forwarded to the supernode that possesses
the same semantic tag as the packet.
3) Prototype development and extensive experimentation
revealed that INF-NDN IoT has optimized the name
length and memory, interest satisfaction rate, hop
count, retrieval time, and energy consumption.
The remainder of this article is organized as follows.
Section II provides some brief background on NDN and
state-of-the-art related works. The design description of the
INF-NDN IoT is outlined in section III. Section IV specifies
the proposed methodology. Section V presents use cases.
Section VI explains the simulation study. Finally, section VII
concludes our work and highlights the future direction.
II. BACKGROUND AND RELATED WORK
A. NDN ARCHITECTURE
NDN is a state-of-the-art technology that aims to overcome
the limitations of the current Internet. It offers a sophisticated
architecture where content is identified through hierarchical
meaningful names rather than its physical location. NDN
has been identified as capable of meeting the requirements
of these newly developed Internet applications [15]. NDN
routers have the ability to cache data while ensuring ade-
quate bandwidth to meet the users’ demands. This makes
the technology highly efficient and reliable in managing
and delivering content to its users [16]. The NDN router is
2VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
PIT FIB
PIT
CS
Interest
Data
Aggrgate Interest
CS
Data
Discard Data
Cache
Drop or NACK
Interest
Data
Upstream
Downstream
No No
Yes Yes No
No
Yes
FIGURE 1: NDN communication process.
endowed with three data structures to attain the NDN stateful
communication architecture, as presented in Fig. 1.
CS behaves like a temporary content store or temporary
cache. CS stores the most recent and frequently asked data in
it. A PIT is served to store the pending request. If data is not
in CS, its name will be investigated in PIT with the incoming
interface. If the name is absent in the PIT, it will be forwarded
to the FIB and a new entry will be inserted in the PIT. FIB
stores the retrieved name announced by the router. It behaves
like a control panel to make proper interest forwarding [17–
19].
B. NAMING IN NDN-IOT
Named-based communication relies heavily on naming as it
directly influences routing, forwarding, and caching mech-
anisms. Researchers have come up with various naming
schemes for NDN-IoT networks. In this section, the state-
of-the-art naming schemes of NDN IoT are summarized
and classified into two types, hierarchical and hybrid, as
described in Fig. 2. The way interest/data are named in the
NDN follows a structure that is arranged in a hierarchy. For
instance, a home page created by hongik may have a name
like /hongik.my/en/main.html, which is similar to URLs but
not necessarily easy for humans to read [20, 21], and also
occupy more space in naming due to repeated line slash
‘/’, such are native naming. Name-based service (NBS) like
[22] follows the same naming structure and uses command
markers to name and access services in a cyber-physical
system (CPS) network. These markers are followed by a
namespace identifier and then the service names, operation
identifiers, and arguments which are delimited by tilde ‘∼’,
e.g.
Content centric networking testbed framework (PHINet)
[23] implemented naming scheme that allows applications
to exchange data over IP using NDN-compliant communi-
cation. It uses UDP to support NDN’s pull-based mechanism
and multitasking. PHINet is implemented using Node.js with
a PostgreSQL database for cloud servers and Nod.js with
Native Naming
Hierarchical
Naming in NDN IoT
Hybrid
NBS Native Naming
PHINET Naming
NDOMUS Naming
NDN-BMS Naming
Homenet Naming
Functional Oriented
Naming
Catagorised Aided Naming
Policy Based Naming
Device & Location-
based Naming
FIGURE 2: NDN IoT Naming Taxonomy.
Sqlite3 for the Android operating system to provide native
UDP support for client-side applications. NDN for smart
home automation systems (NDOMUS) [24] is another nam-
ing scheme used in NDN IoT that supports both sensing and
management operations in a house. It has two sub-namespace
classes: configuration and management, used for network
initialization and management, and task (identified by the
prefix /task,), used for control and monitoring operations.
Building management system (BMS-NDN) naming
scheme designed by [25] uses a hierarchical structure with
a root node that stands for the common prefix for the BMS
namespace. They designed the NDN BMS naming where the
root node represents the common prefix for the university
BMS namespace (/ndn/estc/bms). The building label, located
under the root prefix, is divided into floors based on the
floor number and further subdivided into rooms based on the
room number. Authors in [26] implement the homenet-based
naming scheme in the smart home.
VOLUME 4, 2016 3
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
Hybrid naming schemes have also been developed by
researchers based on specific needs, in which they focus on
a particular need and develop the hybrid design accordingly.
[27] introduced an entropy-based naming scheme in which
the overall name is compressed for efficient name look-up
on the machine level. Performance evaluation includes the
average name compression ratio (ANCR) and distribution
of component compression rate (DCCR). [28] proposed a
naming scheme that is based on PURSUIT architecture.
The name is divided into their parts. Evaluation includes
scheme in the FIB table’s name length and look speed. For
wireless devices, [29] suggested the multilayer multicompo-
nent hierarchical attribute-value (M2HAV) naming method.
The demonstration includes hierarchical-based naming using
prefix labeling and a variable-length encoding technique. The
naming method is divided into four tiers, with a separate set
of characteristics and features used at each level. The name
integrated query (NINQ) [30] architecture, includes three
components—a hierarchical name, a hash-based flat, and a
query-offers a hybrid naming strategy for smart building
using NDN. This naming strategy is promising in terms of
command satisfaction rate (CSR), interest satisfaction rate
(ISR), average delay, energy consumption, and the number of
network packets handled, however, naming and query could
be too lengthy. The naming of physical IoT objects is driven
by using the lightweight named object (LNO) device-based
naming scheme introduced in [31]. This scheme can simplify
programming and enhance device functionality.
C. NLP INTEGRATION WITH NDN
Content names are text-based strings. Therefore, NLP
presents exciting opportunities for efficient communication,
leveraging its specialized functions tailored for textual data.
Despite garnering researcher attention, there is limited work
done in NDN using NLP.
Our target paper, naming and routing scheme for data
content objects in information-centric networks (SICN) [32]
incorporated the naming scheme with forwarding using NLP.
SICN is the only paper to consider semantics in ICN. SICN
suggests that publishers and subscribers should hold dynamic
addresses that can be changed based on their geography in
the network. Also, the names used for dynamic addresses
should convey clearly what kind of information they rep-
resent. Therefore, the authors suggest a three-dimensional
naming system for SICN. User (publisher/subscriber) needs
to mention at least one of these three dimensions when they
label their data. These three dimensions are the publisher ID,
semantic ID, and geographical ID. SICN’s proposed forward-
ing algorithms hold three tables combining the three address
dimensions. The first one is the semantic-ID table that con-
nects the semantic address to the publisher ID address. The
second one is the geo-ID table that connects the publisher
ID address and geographical address, and the third table is
the geo-semantic matches semantic address to geographical
address. SICN router matches the IDs to forward the interest.
If there is no positive match, the router will broadcast the
interest to find the best match. On the contrary, if positive
matches occur, the router forwards the interest accordingly.
However, SICN does not hold a fully semantical forwarding
strategy. The reason is that SICN can also forward data if
the semantic ID is missing in the routing table. In addition,
SICN incorporates geographic ID makes it more suitable for
an IP network than NDN because NDN does not consider
geographical location and SICN lacks completeness with the
constraint of mentioning one dimension by the user.
Another potential work [33] used NLP for naming in an
ICN-based weather monitoring system. A monitoring camera
captures video or images, and AI technologies on the camera
side generate appropriate content names based on the scenes
depicted. In contrast, when users provide names in interest
packets to retrieve content in ICN, the proposed scheme
analyzes priority words from user input sentences on the
user side using NLP. Authors classified words contained in
sentences to various labels with high probability. However,
the authors overlooked the aspect of packet forwarding and
semantic communication based on names.
D. SUMMARY OF RELATED WORK
Many scholarly papers have been published with hierarchi-
cal and hybrid naming in NDN-IoT. The authors designed
schemes according to their appropriate scenarios. However,
the improvement is limited by various factors, such as the
length of the name, errors in names, and overlooking the
potential of names to get forwarding clues. The length of
content names becomes critical, as it can impact the lookup
process in FIB and occupy more memory. This is particularly
crucial in IoT scenarios like data on demand (e.g., military,
healthcare, transportation), where minimizing content access
latency is essential. To address the name length challenge,
further research is necessary to develop naming solutions that
reduce name length without imposing arbitrary word limits
on users.
Moreover, researchers have ignored the ambiguities and
semantical nature of content names. For instance, a requester
requests an interest with the name “document/fiction/”. How-
ever, the forwarder holds a similar-sounding but distinct pre-
fix named “document/friction/”. Both documents have differ-
ent meanings and spellings. Both content names may belong
to different smart environments. Because of the longest prefix
match algorithm in FIB, the NDN forwards interest towards
the interface of “friction" instead of “fiction" since it stops at
the first differing character “F”. This could lead to a semantic
retrieval error since the algorithm prioritizes the longest com-
mon prefix without fully considering the requester’s intent.
Therefore, another notable major loophole in hierarchical and
hybrid naming schemes is the lack of exploration of content
names. Existing naming schemes have left room for how
naming can provide forwarding clues. Table 1 provides an
overview of the most significant literature review.
To address these issues, INF-NDN IoT leverages NLP to
achieve intelligent naming and forwarding. INF-NDN IoT
establishes a correlation between naming and forwarding
4VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
TABLE 1: Summary of Literature Review
Paper Naming
Tech-
nique
Research Gap Major Contribution Communication
Protocol
Semantic
Aware-
ness
Major Limitation
[22] NDN Na-
tive
Integration of SOA in
WSN IoT
Using URNs instead of URLs to
provide a name service-centric ar-
chitecture.
CCNx No Proposed named services require a
consistent naming system within the
CPS. Missing marshalling and un-
marshalling functionality
[25] NDN Na-
tive
Design a new naming
and forwarding strate-
gies using ML
Conception of an intelligent HVAC
control based on deep learning. De-
sign of a new BMS over NDN. Re-
duce energy consumption.
Deep learning
supported
forwarding
information base.
No Overlooks the real-time networking
while implementing the DNN.
[23] NDN Na-
tive
Testbed based research
gaps between health
IoT and NDN
Provide a cheap and easy-to-use
testbed for experimentation with
Health-IoT over the content-centric
network. Architecture that inte-
grates various health sensors such as
those worn on the body
Parser supported
forwarding strat-
egy
No Accessibility issue. Not supported
with additional communication pro-
tocol.
[27] Hybrid
naming
Entropy-based naming Encoding the original names for ef-
ficient communication and look up
Native No Lack of update policy for con-
tent names. Proposed compactTrie
adopts Pointer which is a further
waste of memory.
[28] Hybrid
naming
(Pursuit
based)
Long length of names,
difficulty of finding
unique content
in attribute based
naming and naming
complexities.
Pursuit based NDN architecture for
IoT smart city.
Publish-
Subscribe
Internet Routing
Paradigm
(PSIRP) based
protocol
No Too much geographic information is
incorporated in name design (loca-
tion dependent).
[29] Hybrid
naming
(M2HAV
based)
Naming basic issues,
such as globally
unique, persistent and
secure.
Self-certifying names to achieve a
standardized naming scheme
Tree-structured
supported
categorical
based forwarding
No Name and location is merged in pro-
posed naming.
[32] Hybrid
naming
(IDs
based)
Naming and routing Classifying data into the four types
and classifying subscriber request
into four classes, afterwards naming
and routing accordingly
ID based forward-
ing
Yes (par-
tially)
Broadcasting if IDs are not matched
(more delay, more energy, etc) and
high length of names.
[33] Hybrid
naming
(ML-
based)
Lack of design espe-
cially supported for the
weather monitor sys-
tem
New naming structure to be used for
only weather IoT sensor or monitor-
ing camera applications
NDN native No Limited to weather monitor condi-
tion.
with the help of semantic tags. The key idea is to explore
the FIB names associated with various smart areas within
an NDN-based IoT framework. This exploration allows the
proposed methodology to identify the primary domain that
each smart area handles, such as libraries or classrooms. Sub-
sequently, upon receiving an incoming interest, INF-NDN
IoT analyzes its name and directs it to the appropriate smart
area that corresponds to its request. For instance, an interest
like “/Author/musa-raza/books/friction" would be routed to
the smart library within the NDN-IoT framework. In essence,
INF-NDN IoT classifies NDN-IoT data into distinct semantic
or contextual categories and facilitates packet forwarding
based on the semantic category. To our knowledge, we are
the first to use NLP to obtain the forwarding clue (semantic
tag) from the NDN name and perform semantic forwarding
in NDN IoT.
III. DESIGN COMPONENT OF PROPOSED MODEL
The design component of the proposed scheme involves the
three types of forwarder nodes to achieve the desired objec-
tives. It is pertinent to accentuate that designated forwarder
nodes are resource-rich.
A. PRINCIPAL NODE
Principal node performs NLP-based naming tasks and finds
the set of supernodes in the network. The principal node as-
signs the semantic tags to names and supernodes. Moreover,
the principal node is connected to supernodes as a simple tree
topology. It is to be noted that the selection of the principal
node is out of the scope of this work.
B. SUPERNODE
The interconnection of supernodes ensures that every smart
area within the smart campus is encompassed by a cor-
responding supernode. Supernode holds a shared complete
database (discussed in detail in the last of the current section)
to assist local FIB. Every supernode is aware of the entire
topology including contextual tags of the other supernodes
in the network such as [34, 35]. However, the scalability of
supernodes is limited to smart campus.
C. ORDINARY NODES
These nodes are further child nodes of supernodes. One
ordinary node can be connected to more than one parent
supernode. The ordinary node’s FIB contains name prefixes
only about parent supernodes.
D. COMPLETE DATABASE (CDB)
CDB is maintained at each supernodes [36]. CDB is similar
to link state database (LSDB) in name link state routing
(NLSR) [37]. However, in contrast to traditional FIB which
contains name prefixes and the corresponding outfaces, the
VOLUME 4, 2016 5
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
LSA-
LSA-
LSA-
LSA-
Name Next Hop
Set
NA-1.1
Name Next Hop
Set
NA-2.1
NA-3.1
Name Next Hop
Set
NA-1.1
CDB-Based FIB in CDB-Based FIB in
CDB-Based FIB in
LSA-
LSA-
LSA-
LSA-
Name Next Hop
Set
NA-1.1
NA-3.1
Name Next Hop
Set
NA-2.1
NA-3.1
Name Next Hop
Set
NA-1.1
NA-2.1
CDB-Based FIB in CDB-Based FIB in
CDB-Based FIB in
( A) ( B )
FIGURE 3: Updation of CDB-based FIB.
CDB-based FIB includes all the name prefixes served by
every supernode in the network. To this end, ordinary nodes
and supernodes send link state advertisement (LSA) to nodes
in their vicinity, at regular intervals to keep them updated
[38].
To understand CDB-based FIB of supernodes, Fig. 3 il-
lustrates the scenario where three supernodes with semantic
tag N1,N2, and N3are producers of prefixes NA-1.1, NA-
2.1, and NA-3.1 respectively. At the initial stage, all nodes do
not have the prefixes served by each other. As illustrated in
Fig. 3 (A), all supernodes exchange LSAs with neighboring
supernodes when the network becomes operational. The LSA
of each supernode carries prefixes present in its FIB. When
N1receives LSAs from N2and N3, it updates its CDB-
based FIB with the information about N2’s prefix (NA-2.1),
N3’s prefix (NA-3.1) and sets the next hop from which LSA
of N2and N3was received. Same as N2and N3updates
its CDB-based FIB with the information about N1’s prefix
(NA-1.1) and sets the next hop. However, N2and N3are
still unaware of prefixes of each other. When all supernodes
again exchange LSA, as shown in Fig. 3 (B), N2and N3
receive LSA from N1, receive prefixes NA-2.1 and NA-3.1
respectively. As prefixes are received from LSA of N1, as a
result, N2and N3set the next hop N1for prefixes NA-2.1
and NA-3.1.
All supernodes exchange LSA at regular intervals. Conse-
quently, any modifications made to the CDB of a single for-
warder are automatically disseminated to all other forwarders
within the network, ensuring synchronized data across the
network. The method for calculating the time required to
exchange the LSA is beyond the scope of our work.
IV. PROPOSED MODEL
INF-NDN IoT is deployed on smart campus. The selection
of a smart campus for semantic communication is motivated
by the diverse IoT configurations present in a smart cam-
pus, such as smart labs and smart classrooms, etc. Each
smart environment is tailored to manage specific activities;
for instance, a smart library handles library-related traffic,
and a smart classroom manages classroom traffic. Hence,
predicting the destination from the interest name is logical
and practical within the smart campus. Figure 4 demonstrates
the flow and architectural diagram provided by the INF-NDN
IoT.
In this illustration, the principal node receives two types
of data from within the smart campus. The first type (FIB
names of supernode, briefly discussed in section IV.B.3) is
represented by a green dotted arrow, presenting the flow of
semantic tag assignment to supernodes. The second type is
depicted by a dark navy blue arrow, signifying the seman-
tic tag assignment to packets (briefly discussed in section
IV.A.4). Arrows of sky blue color present data flow from the
outer NDN network. For the outer NDN network communi-
cation, INF-NDN IoT performs formatting (briefly discussed
in section V) after receiving and sending of interest and
data packets, respectively. The rationale behind this is outer
NDN network’s packet namespace designs may be different.
Mainly INF-NDN IoT is comprised of two subsections.
1) NLP-based intelligent naming with semantic tag.
2) Semantic tag-assisted supernode-based forwarding.
A comprehensive description of the aforementioned con-
stituents is provided as follows.
A. NLP-BASED INTELLIGENT NAMING WITH SEMANTIC
TAGS
INF-NDN IoT develops a namespace design to enable com-
munication between NDN IoT networks in the smart campus.
Each node can forward and receive interest/data to other
nodes according to its namespace design. It is to be noted
that IDs presented in the namespace designs do not violate
the basics of NDN. The rationale for introducing IDs in the
namespace is mentioned along with detail. A comprehensive
6VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
List of FIB
names
Interest/Data
name
Apply NLP to:
Optimize the name, removing text based naming errors
and explore the name to calculate the semantic tag
Reformatting of Interest packet
Label the semantic
tag to super node,
such as smart
library, etc.
Add the semantig
tag to packet's
optimized name
Add the semantig
tag to packet's
optimized name
Forward the packet to super having
same semantic tag as packet name
Super Nodes Send Data
1
2
3
1
2
3
4
3
,
4
5
1
7
Run the algorithm for selection
of super node
recieves packet
(NDN network or Super
node)
Start
Outer NDN network
Super Node
Super node removes the semantic tag
from namespace design, arrenge the
data
Data type
Data for
Reformatting of
Datapacket by principal
node
Send by reverse path
Other super nodes Outer NDN network
56
Send by reverse path
END
END
Interest packet
2
if name pre fix of interest
not found in CDB-FIB,
forward the packet to
principa l node for NLP
operation s
Super Node
Smar IoT Area
Smart IoT Area
Outer NDN
network
Super Node
Provide the FIB names to
principal node to aquire
the semantic tag as
identity of super node
Principal
Node
Smart IoT Area
Super Node
Principal
Node functions
For Naming
For Forwarding
After selection
of supernode
FIGURE 4: INF-NDN IoT architecture.
description of the namespaces introduced by the INF-NDN
IoT is provided as follows.
1) Namespace Design of Ordinary Node
The namespace design for ordinary nodes, as illustrated in
Fig. 5, includes several key components. The first compo-
nent pertains to the unique ID of the ordinary node. The
supernode uses this ID to distinguish the current node from
other ordinary nodes and their produced data as well , such as
[39–41] used the nodes ID in the context of NDN. The INF-
NDN IoT does not use the ordinary node ID for supernode-
to-supernode or supernode-to-principal node communication
due to the data-oriented architecture of NDN. The second
component relates to the supernode semantic tag (section
IV.B.3), facilitating intersupernode forwarding. The third
component represents the type of data generated by the IoT
Format of ordinary node Interest messege:
Format of ordinary node Data messege:
/ node-ID / super-node semantic tag / type | non-query
/ node-ID / super-node semantic tag / customizedretrieval | Data
/ node-ID / super-node semantic tag / type | query
customize retrieval
FIGURE 5: Namespace of ordinary node.
devices (e..g., temperature, pressure, humidity, etc.) or query
(section IV.A.4) of a random user. This namespace design is
used to communicate between the child ordinary node and
the parent supernode.
2) Namespace Design of Supernode
To support the roles of supernodes effectively, a supernode
requires a well-defined namespace that incorporates the spe-
cific type of content produced under its vicinity. INF-NDN
IoT uses two distinct namespace designs for supernodes, as
illustrated in Fig. 6 and 7. The namespace design presented
in Fig. 6, is only leveraged to acquire the semantic tags
to supernodes. The first component in the interest packet’s
namespace denotes a supernode ID, such as [42–44] lever-
aged from nodes ID in NDN.
The second component is a list of content names present
in the FIB of supernode. The principal node receives the
interest, performs NLP-based operations (section IV.A.4),
Format of super node Interest messege:
Format of super node Data messege:
/ super-node ID / list of names
/ super-node ID / super node semantic tag | Data
FIGURE 6: Namespace of supernode - A.
VOLUME 4, 2016 7
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
and assigns a semantic tag to each corresponding supernode
in exchange.
Once semantic tags are assigned, the rest of the commu-
nication is performed by the namespace design of Fig.7.
This namespace design serves as the backbone of the smart
campus network. Fig. 7 comprises key components of the
namespace design when the supernode communicates with
other supernodes and principal nodes after acquiring the
semantic tags. The first component in Fig.7 is a semantic
tag. The second component is the type of data that can be
a query or non-query. It is to be noted that semantic tags are
not permanently assigned to supernodes. To support the data-
oriented architecture of NDN, semantic tags can be changed
anytime with contextual variation in network traffic.
Format of super node Interest messege:
Format of super node Data messege:
/ super-node semantic tag / type | non-query
super node semantic tag / customizedretrieval | Data
/ super-node semantic tag / type | query
customize retrieval
FIGURE 7: Namespace of supernode - B.
3) Namespace Design of Principal Node
The namespace design for a principal node includes several
components. As illustrated in Fig. 8, the first component can
be either a supernode ID or a supernode semantic tag.
Format of principal node Interest messege:
Format of Principal node Data messege:
/ super-node ID / type | list of names
/ super-node ID / semantic tag | Data
/ super-node semantic tag / type | query
customize retrieval
/ super-node semantic tag / query | Data
FIGURE 8: Namespace of principal node.
The supernode ID is only essential for acquiring semantic
tags based on the shared list of FIB names. Conversely, the
supernode semantic tag serves as a forwarding clue. The
second component is either the list shared by supernodes or
the name in the form of the query. The principal node replies
with a requested data packet accordingly. The principal node
uses its namespace design to only send and receive packets
with supernodes. For outer communication, INF-NDN IoT
assumes that the principal node only receives interest packets
from outside the smart campus network. On receiving interest
from outside the smart campus, the principal node applies
NLP to change the format of incoming interest according
to its namespace and forwards it to appropriate supernodes.
After receiving the data from the supernode, the principal
node reformats the data packet into the original format (i.e.,
as per reprieved Interets packet format) and sends it back by
reverse path.
4) Query Optimization and Semantic Tag to content names
In INF-NDN IoT, optimization refers to reducing the length
of content names and removing equities from them. There
are no restrictions on the word count for queries in INF-NDN
IoT, as all forwarder nodes are resource-rich. The workflow
and major steps for the query optimization is outlined in Fig.
9 followed by Fig. 10. To clarify the steps involved, INF-
NDN IoT focuses on a specific use case: fetching the bus
schedule for Hongik University from the Jochiwon station
to the Sejong campus. This scenario employs the following
namespace for query execution.
Name =“/students/bustiming?
from$JOCHIWON$to+hongik?Sejong
_campus”
After performing segmentation, cleaning process, normaliza-
tion, abbreviation, stemming, and stopword removing steps,
the query is optimized as depicted below and does not include
extra characters in it, as described.
Optimized_name =“studentbustimejochiwon
hongiksejong”
Afterward, the count vectorizer technique [13] is utilized for
the probabilistic formulation of the query. This method trans-
forms the query into vector form, facilitating the subsequent
application of NLP algorithms. The semantic score of the
optimized query is calculated to understand the context.
To this end, INF-NDN IoT implements the NLP technique
called latent dirichlet allocation (LDA) [45, 46] to generate
and allocate a semantical tag to each NDN name and supern-
ode. LDA is a probabilistic unsupervised classification tech-
nique in topic modeling [47]. Through an iterative process of
analyzing the co-occurrence of names in the NDN naming,
this semantic model endeavors to identify the contextual tag
and represent each document (content names) as a combina-
tion of various contextual tags with corresponding weights.
For the implementation of LDA, we define a set of
documents as content names and words as the topic or
semantic tag, where discrete tag distributions are mapped
from LDA. Mathematically, given a vocabulary with N
number of tags, document D=i1, i2, i3, . . . , in. for
each name iin D, we perform the following procedure:
Given the joint distribution parameters αand βcontrol tags-
names distribution and individual word-to-tag distributions
in LDA over the random variables ix, jx, θx, φTto denote
the tag-name relationship, are given by:
P(ix, jx, θx, φT|α, β) = P(θx|α)P(φT|β)
Nx
Y
n−1
P(jx,n|θx)P(ix,n |φjx,n )(1)
Embedding over θxand summing over jx,n, the marginal
distribution of an NDN name can be computed:
8VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
Preprocess
Segmentation
Cleaning process
Normalization
Abbreviation
Stemming
Stop words removal
Probabilistic presentation
Word vector conversion
Semantic tag
Dividing NDN name into
its individual words
Removing noise, such as special
characters and punctuation from
text data to enhance its quality
converting name into a
standard format, such as
converting all words to
lowercase Allows the system to understand
and interpret text containing
abbreviations.
Reducing words to their root
form, which involves removing
suffixes and prefixes Filtering out common words,
such as "the" and "and," from text
data to improve the accuracy of
text analysis
Enabling machines to analyze
the frequency and
relationships between words
in a given context
Transforming text data into a
mathematical representation,
such as a vector
Finding similarity between
vectors, extract meaning, and
assign a tag
FIGURE 9: NLP-based tasks in intelligent naming.
Students’ bus timing from JOCHIWON to hongik
Sejong_campus
Students’ bus timing from JOCHIWON to hongik
Sejong _ campus
Students bus timing from JOCHIWON to hongik
Sejongcampus
Studentsbustimingfromjochiwontohongiksejongcam
pus
Studentsbustimingfromjochiwontohongiksejong
Studentbustimefromjochiwontohongiksejong
Studentbustimejochiwonhongiksejong
Ngram/TF-IDF/Countvectorizer
V1,V2,V3,V4,....,Vn
<Transportation>Studentbustimejochiwonhongiks
ejong
Content Name
Semantic tag
Segmentation
Cleaning process
Normalization
Abbreviation
Stemming
Stop words removal
Probabilistic
presentation
Word vector
conversion
FIGURE 10: An example of finding a tag from the query.
P(ix|α, β) = ZθxZφT
P(θx|α)P(φT|β)
Nx
Y
n=1 X
jx,n
P(jx,n|θx)P(ix,n |φjx,n )
×dθxdφT
(2)
At the last, by computing the marginal probability of each
query, the probability of constructing the set of documents is
established through the following method:
P(D|α, β) =
X
Y
x=1 ZθxZφT
P(θx|α)P(φT|β)
Nx
Y
n=1 X
jx
P(jx,n|θx)P(ix,n |φjx,n )
×dθxdφT,
(3)
The LDA model is learned using Gibbs sampling, and
subsequently, each instance is represented using topic distri-
butions. To assign a new label (contextual tag in our case) to
an Nn, we utilize a Gibbs sampler by sampling:
P(jk=T|jk−1, d, i)∝nit +β
Pv(nvt +βv)(ndt +α)(4)
The counts of the contextual or semantic tag with words i
or in Dare denoted by nit and ndt respectively. The hyper-
parameters αand βremain the same as before. Following the
above-mentioned intelligent naming procedure, every query
would be optimized and semantic tags would be assigned to
them. The semantic tag summarises the contextual nature of
the NDN name into a single word. Along this, semantic tags
are assigned to supernodes in the network. The “Transporta-
tion” tag has been assigned to the query in the above-said
example. This tag describes the forwarding clue for the NDN
interest packet, as well (explained in section V). The seman-
tic tags provide a concise representation of the semantical
nature of the NDN packets and help to reach their intended
destinations or supernodes within the IoT network. As shown
in Fig. 11, different content names have been optimized and
related semantic tags are assigned. All the intelligent naming
operations are implemented on the principal node.
VOLUME 4, 2016 9
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
NDN Names
Students’ bus timing from JOCHIWON to hongik
Sejong _ campus
best goals-in university semester gala in 2023
Ph.D. student’s results of Professor Kim
marking sheet of Network Performance Analysis
students honda cars display
science fiction based latest books
water level of garlic slot 4
Student joining data to social societies
today organic level slot 15
Semantic Tags
<Agriculture>
<Transportation>
<Sports>
<Exams>
<Exams>
<Parking>
<Library>
<Agriculture>
<Admin Block>
FIGURE 11: Examples of tags corresponding to the content
names.
B. SEMANTIC TAG-ASSISTED SUPERNODE-BASED
FORWARDING.
Packet dissemination is accomplished through a semantic
tag-assisted supernodes-based forwarding algorithm in the
proposed model. Before diving deep into the explanation, we
like to explain why we need a set of supernodes in INF-NDN
IoT to forward the packets, how does the principal node select
and assign the semantic tag supernodes?
1) Idea Behind the Supernodes
The idea behind supernodes is to identify the smallest set
capable of maintaining the network’s connectivity and ensure
its connectivity with the least possible number of connec-
tions. To understand the forwarding concept of supernodes,
we take a network graph of multiple nodes, as shown in Fig.
12 (A). Upon implementing the selection of the supernodes
(algorithm 1), the nodes interconnect each other in such a
way that each one establishes communications links with
other nodes. This results in a connected subgraph, thereby
ensuring comprehensive network connectivity. These specific
nodes are referred to as supernodes within the context of INF-
NDN IoT. As shown in Fig. 12 (b), the set of supernodes
includes nodes A, B, C, D, and E.
2) Selection of Supernodes
The network initiates randomly for the selection of supern-
odes in the smart campus. The principal node runs algorithm
1 to find the supernodes in the network.
The algorithm’s execution comprises two fundamental
steps: initialization and supernode selection. A detailed de-
scription of algorithm 1 is presented as follows: The al-
gorithm operates on an input graph (network topology),
N N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
(a) Intennectivity of nodes before selection of supern-
odes.
D E
CB
A
(b) Intennectivity of nodes after selection of supern-
odes.
FIGURE 12: Comparison of interconnectivity before and
after selection of supernodes
Algorithm 1 Selection of supernodes
1: Input: Graph=(V, E ), vertex set Vand edge set E
2: Source - vertex ←s∈V
3: Auxiliary variables and functions:
4: LSS - list, Covered - list, Neighbours(v)for all v∈V
5: Output: LSS - list
6: Initialization: Covered - list = (s), Covered - list = Φ
7: Begin d- selection
8: While |Covered - list|<|V|
9: Select a vertex r∈Covered - list and
10: r /∈LSS - list such that rhas the maximum
11: neighbors that are not in Covered - list.
12: LSS - list = LSS - list U(r)
13: for allu∈Neighbours(r) and U /∈Covered - list
14: Covered - list = Covered - list U(u)
15: end while
16: return: LSS - list
17: End d- selection
18: Terminate
Graph=(V, E), where V is the set of vertices (nodes) and E is
the set of edges (links) connecting them. The principal node
initiates by selecting random “s” as the source vertex line (2).
Afterward, the principal node creates two lists: “LSS - list”
(linked spanning set) to store supernodes and “Covered-list”
to keep track of covered nodes (line 4). The main goal is to
find key nodes, called supernodes, that can improve network
connectivity.
The algorithm begins with the source node in the “Covered
- list” and an empty “LSS-list” (line 6). It enters a loop that
continues until all nodes are covered (line 8). In each loop,
it chooses a node from the “Covered - list” that hasn’t been
included in the “LSS - list” and has the most unconnected
10 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
to
to
to
FIGURE 13: Semantic tag-assisted supernodes-based forwarding.
neighbors (lines 9-11). This node is added to the “LSS -
list” as a supernode (line 12). Additionally, the algorithm
adds the neighbors of the chosen node to the “Covered - list”
if they’re not already covered, expanding network coverage
(lines 13-15). The loop keeps running until all nodes are
covered, ensuring complete network coverage. Finally, the
algorithm returns the “LSS - list”, which contains the selected
supernodes (line 16).
It is to be noted that the updation time of supernodes is
application-specific. After every updation, fresh supernodes
and corresponding tags would be assigned. Once supernodes
are selected, NLP-based algorithms are employed to assign
semantic tags to these nodes, as described in the next section.
Only supernodes need to update the CDB and send LSA to
other supernodes to keep them updated. As a result, commu-
nication costs for synchronization are significantly reduced
(briefly discussed in section VI.B).
3) Semantic Tags Assignments to Supernodes
Every supernode initiates traffic generation and forwards a
packet to the principal node. This packet comprises a list
of FIB entries of the respective supernode, as shown in Fig.
13 where orange arrows represent the principal node Npto
the supernode Nsconnectivity and blue arrows represent
the supernode to supernode connectivity. The principal node
performs the NLP-based LDA to names in the shared list, to
understand the contextual nature of traffic in nodes (as briefly
discussed in section IV.A.4). As a result, semantic tags are
generated and sent back to supernodes in data packet format.
In essence, every supernode is enriched with a semantic
tag. As a result, the combination of tags and supernodes
has categorized the network into different smart areas. This
enables NDN packets to be forwarded based on semantic
context, thereby potentially optimizing the routing paths.
Leveraging this semantic-based forwarding algorithm, the
INF-NDN IoT significantly minimizes the latency required
to retrieve the data, as compared to issuing multiple interests
for each content object. For instance, if the supernode 1 of
the < smartlab > in smart campus generates the majority
of lab-related requests as compared to other nodes. The
principal node assigns < smartlab > as a semantic tag to
that supernode accordingly.
The same applies to the < smartdormitories > su-
pernode 2, which handles the network traffic associated with
dormitories-related Interest/Data.
However, it’s important to note that any supernode is not
exclusively limited to the corresponding semantic tag. The
principal node can assign multiple tags to single super node.
V. USE CASES SCENARIO
We formulate four distinct use case scenarios to elucidate the
concept of tag-assisted supernodes-based forwarding.
• Usecase A: Principal node to supernode communication
(query).
VOLUME 4, 2016 11
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
• Usecase B: Supernode to supernode communication.
• Usecase C: IoT devices communication (non-query).
• Usecase D: Assigning multiple tags to a single supern-
ode.
These use case scenarios serve as practical illustrations to
showcase the effectiveness and applicability of the semantic
tag-assisted supernodes-based forwarding approach in IoT
scenarios. We also assume push-based communication in
INF-NDN IoT, which supports IoT devices to push the data
Algorithm 2 Semantic Tag assisted supernodes based for-
warding
1: Initialize the network randomly;
2: Nprun algorithm 1 for selection of Ns;
3: Apply NLP based LDA to Nnof each Ns;
4: Assign Tto each Ns;
5: Initialize the CDB;
6: Use case: A
7: ForPnat Np
8: Check Ft;
9: If Cnof Pnnot found
10: Apply NLP-based text analysis;
11: Optimize the Nn;
12: Perform NLP-based semantic analysis;
13: LDA assign a Tto the Nn;
14: Forward the Pnto the relevant Ns;
15: For relevant Ns
16: Removes the Tfrom the Cn;
17: If Packet is Pi
18: Check CS and send the Pdto the reverse path;
19: Else Packet is Pd;
20: Check the PIT and respond correspondingly;
21: EndIf
22: EndFor
23: EndIf
24: Repeat the process for each Pnat the Np;
25: EndFor
26: Update the CDB;
27: Use case: B,C
28: ForPnat the Ns
29: For each Pi
30: Check the Ft(CDB-based);
31: If not found in Ft(CDB-based)
32: Send the packet to Np;
33: Repeat process from line 7 to 26;
34: Else Found in Ft(CDB-based)
35: Forward the packet to the relevant Ns;
36: EndIf
37: EndFor
38: For Packet is Pd
39: Send the Packet to the Np;
40: Repeat process from line 7 to 26;
41: EndFor
42: For packet is Ps,s
43: Check if Ps,s is Interest/Data and responds accordingly
44: EndFor
45: Repeat the process for each Pnat the Ns;
46: EndFor
47: Update the CDB;
48: Terminate.
packet into the network without interest, such as [48]. The
forwarding strategy for the use cases is outlined in algorithms
2. A comprehensive list of all notations employed in algo-
rithm 2 can be found in Table 2.
TABLE 2: Description of notations
Notations Description
PnNDN packet
CnContent name
FtForwarding table (FIB)
NpPrincipal node
NsSupernode
NoOrdinary node
TSemantic or contextual tag
PiInterest packet (Query)
PdData packet
CDB Complete database
Ps,s Packet of IoT devices (non-
query)
A. USECASE A: PRINCIPAL NODE TO SUPERNODE
COMMUNICATION
In this case, the principal node receives an interest packet
from an external disjoint network source. A stepwise descrip-
tion of interest and data packet exchange shown in Fig. 14 is
provided as follows
Step 1: The interest packet arrives at the prin-
cipal node “P” and carries the name of content
to be retrieved, “available/freehostels$for?
students_of/Ph.D”.
Step 2: Principal node “P” receives the interest. As this
interest is received from the outer disjoint NDN network,
therefore reformatting of such interest packets is performed.
To this end, query optimization is performed and LDA is
applied to assign the semantic tag to the optimized query.
C
A
5
S
G
//
*
*
*
*
P
*
*
*
*
*
*
*
D
Principal Node
Super Node
Ordinary Node
Data
Interest
Update
1
2
34
5
6Internet
Dormitories/availablehostelsphdstudent
availablehostelsphdstudent/8
Principal node and supernode connectivity
Supernode to supernode connectivity
Supernode to ordinaryrnode connectivity
FIGURE 14: Usecase A.
12 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
Afterward, the principal node “P” reformat interest according
to the format of INF-NDN IoT, which is “Dormitories |
availablehostelphdstudent”, where “Dormitories”
is semantic tag and “availablehostelphdstudent”
is optimized query. In Fig. 14, NLP-based interest is pre-
sented in the green box. At this point, the principal node “P”
forwards this interest to the corresponding supernode with
the same semantic tag as the interest. To this end, supernode
“D” is that corresponding supernode. It is to be noted that if
the semantic tag of the interest name does not match with
the semantic tags of supernodes, the principal node drops the
interest.
Step 3: Supernode “D” receives the interest from the prin-
cipal node and checks the local cache, if data is not found,
then supernode “D” checks the FIB and forwards the interest
to its producer child ordinary node whose ID sign is“//”. We
assume that ordinary nodes are sharing the content names
with parent supernodes by LSA (as mentioned in section
III.D).
Step 4: Ordinary node “//” receives the interest from
supernode “D” and responds to the data packet against re-
quested interest.
Step 5: Supernode “D” receives the data packet from
ordinary node “//”. It may cache the data into its CS and
forward it back to the principal node “P”. As all supernodes
share the common CDB, supernode “D” sends LSA messages
to update the neighbor supernodes about this change in CDB,
as presented by purple arrows in Fig. 14.
Step 6: “Dormitories | availablehostels
phdstudent | 8” is received from supernode “D” where
“8” is the requested content (i.e., available hostels phd
students), as presented in the red box in Fig. 14. At this point,
the format of the data packet is according to INF-NDN IoT, as
presented in the red box. Now, the principal node “P” again
converts the format of the data packet into the same format
in which its interest was received and sends it back to the
internet by reverse path.
It is to be noted that data “8” is included in the namespace
to only present the workflow. In simulations, data is not
placed within the namespace field of the data packet. The
rationale is that the data packet format has a separate field for
the placement of data.
B. USECASE B: SUPERNODE TO SUPERNODE
COMMUNICATION
In this case, multiple scenarios demonstrate efficient inter-
supernode communication within the smart campus IoT net-
work where the supernode may fetch the data from another
supernode.
Step 1: As explained in Fig. 15, where the user connected
to the child ordinary node whose ID sign is “#” requests an
interest packet. The name of interest is “od# / sports
| sectionBstrength ”, where “#” is the ID of the
ordinary node, “sports” is the semantic tag of parent su-
pernode and “sectionBstrength” is the query, as shown
in the red box in Fig. 15.
C
A
5
S
G
**
@
+
#
p
&
*
*
*
*
*
!!
$
D
1
2
3
4
5
6
Principal Node
Super Node
Ordinary Node
Principal node and supernode connectivity
Supernode to supernode connectivity
Supernode to ordinaryrnode connectivity
Data
Interest
Update
od# / sports / sectionBstrength
od# / sports / availablehostelsphdstudent / 23
sports / sectionBstrength
sports /
availablehostelsphdstudent / 23
step
step
od$ / sports / availablehostelsphdstudent / 23
FIGURE 15: Usecase B.
Step 2: Supernode “S” receives the interest. If the
requested data in not present in the cache of supern-
ode “S”, then supernode “S” checks its CDB-based FIB.
If the longest prefix algorithm cannot find the query
“sectionBstrength” in CDB-based FIB, supernode
“S” forwards interest to principal node “P”, repeating steps
2 to 5 from use case A. On the contrary, if the longest
prefix matches the name, supernode “S” finds the query
‘sectionBstrength” in CDB-based FIB, and data is
available on supernode “C’. Supernode “S’ removes the
ordinary node ID “#” from the format of the packet. It is to be
noted that an ordinary node ID is not required for supernode-
to-supernode communication. The rationale for removing
ordinary node ID is to obey the data-oriented nature of
NDN. Supernode “S’ forwards the interest to supernode “C’.
Supernode “C” receives the interest and checks the local
cache.
If requested data is not present in the cache of supernode
“C”, then supernode “C” forwards the request to producer
child ordinary node “$”.
Step 3: Ordinary node “$” receives the interest from parent
supernode “C” and responds to the data packet against the
requested interest.
Step 4: Supernode “C” receives the data packets, caches
into CS, and sends LSA to neighbor supernode “A” to keep
CDB updated.
Step 5: Supernode “C” removes the ordinary id “$” from
the format and forwards the data packets by the reverse path
to supernode “S”.
Step 6: Supernode “S” caches the data to its CS and
forwards the data packet to ordinary node “#”, where
the requester is attached. Data packet “od# / sports
| section strength | 23” is received by ordinary
node “#” where 23 is the only presentation of data but placed
VOLUME 4, 2016 13
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
C
A
5
S
G
**
*
*
#
p
*
*
*
*
*
&
*
$
D
1
2
3
4
5
6
Principal Node
Super Node
Ordinary Node
Principal node and supernode connectivity
Supernode to supernode connectivity
Supernode to ordinaryrnode connectivity
Data
Interest
Update
agri /water_level_DHT22_25
agri /water_level_DHT22_25 / 13
FIGURE 16: IoT devices communication.
into another field of the data packet.
C. USECASE C: IOT DEVICES COMMUNICATION
Within a smart campus, an array of IoT devices are usually
deployed across various zones, such as the smart library and
smart agriculture. Consider a use-case scenario (as shown in
Fig. 16) where two distinct sensing devices want to commu-
nicate with each other.
Since their communication is solely based on exchanging
digital readings, such as temperature and humidity, without
any additional context, the generated names are context-free,
as presented in Table III. Therefore, in this instance, non-
query-based communication will take place.
TABLE 3: Interpretation of non-query in INF-NDN IoT
Non-query
Names
Description
water.level_-
DHT22_25
Select the water level under the sensor DHT22, where
values are higher than 25
humi.level_-
DHT11_10
Select the humidity level under the sensor DHT1,
where values are less than 10
Tem.level_-
TP22_10C
Select the temperature under the sensor TP22, where
values are higher than 10
A stepwise description of interest and data packet ex-
change shown in Fig. 15 is provided as follow:
Step 1: As explained in Fig. 16 (smart agriculture area in
smart campus), child ordinary node “&” of supernode “A”
generates and forwards the interest to parents supernode “A”
Step 2: Supernode “A” receives the interest from ordinary
nodes, removes the ordinary node’s id “&”. In essence, re-
ceived interest’s name is changed to format of supernode
to supernode communication, “agri /water_level_-
DHT22_25” where, “agri” is the semantic tag of parent
supernode and “water_level_DHT22_25” shows the
type of requested data (non-query) related to wired based
sensor “DHT22_25”, as presented in the green box in Fig.
16. To this end, supernode “A” does not forward this interest
to the principal node “P”. Supernode “A” don’t have the
requested data in the cache and check CDB-based FIB with
the help of the longest prefix match algorithm. According to
the CDB-based FIB, data against requested content is placed
on supernode “G” which is one hop away. Supernode “A”
forwards this interest to supernode “D”.
Step 3: Supernode “D” receives the interest, also doesn’t
have the requested data in the cache, and checks the CDB-
based FIB. Data against requested content is placed on su-
pernode “G” which is now the neighbor node of supernode
“D”. Supernode “D” forwards the interest to supernode “G”.
Step 4: Supernode “G” receives the interest packet and
either checks whether the content has already been generated
or senses the environment in real-time. Afterward, supernode
“G” stores the data in CS, inserts an entry in CDB-based FIB,
and sends back the data to supernode “D”.
Step 5: Supernode “D” receives the data, stores it in CS,
updates the CDB, and forwards the data to supernode “A”.
Now, CDB is again updated. Therefore, supernode “D” also
sends LSA to update supernode “S” about the change in
CDB.
Step 6: Supernode “A” receives the data, stores it in CS,
updates CDB, sends LSA to neighbor supernode “C”, and
forwards the data to child ordinary node “&”, as presented in
the red box in Fig. 16.
D. USECASE D: ASSIGNING MULTIPLE TAGS TO A
SINGLE SUPERNODE
INF-NDN IoT considers random and dynamic traffic gen-
eration. However, it is important to note that the location
of supernodes, ordinary nodes, and principal nodes remains
static. When a consumer/producer switches to a different
supernode or ordinary node and initiates communication, the
entire network is promptly informed due to CDB. For in-
stance, if students associated with the < smartclassroom >
tags are switched to < smartlibrary > supernode, it
would imply that a single supernode is managing traffic
for two distinct tags. Therefore, the principal node will
add one more tag < smartclassroom > to supernode
< smartlibrary >. Now supernode < smartlibrary >
will deal as dual tag < smartclassroom_library >. In this
situation, all classroom-related traffic would be forwarded to
< smartclassroom_library >.
VI. EXPERIMENTAL ANALYSIS
This section aims to assess the tradeoffs of the INF-NDN
IoT system by conducting a comprehensive simulation study.
We aim to compare its performance with the state-of-the-
art scheme SICN [32] (briefly discussed in section II.C). We
begin by describing our simulation setup and subsequently
present the results obtained from our simulations.
14 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
A. EXPERIMENTAL SETUP
In our simulation setup, we implement INF-NDN IoT in
Python using python-NDN [49], pyndn2 [50], and NLTK
[51]. Ten different semantic tags are obtained from LDA.
LDA is trained over 1500 NDN IoT content names that
are relevant to the target supernode. The network model
includes a total of sixteen nodes. All the nodes follow wired-
based communication model. Five out of sixteen nodes are
supernodes converged due to CDB. Four out of sixteen nodes
are randomly selected as consumer nodes and two nodes
are publishers. The transmission range of ordinary nodes is
limited to one hop (i.e., to the supernode they are associated
with). The principal node is limited to supernodes in the
network(i.e., to the supernodes they are associated with). The
principal node is located in the center of the network and is
reached by its supernodes. In simulation study, we considered
the following evaluation metrics.
1) Name Length: Total characters in the names of inter-
est/data packets.
2) Name Memory Utilization: Memory utilization by
names of interest/data packet.
3) Retrieval Time: Total amount of time the interest
packet takes to successfully reach the producer, time
taken by the producer to process, and Data packet to
reach back to the requester.
4) Number of Hops: It refers to the count of intermediary
nodes that packets traverse between the provider and
consumer.
5) Intrest Satisfaction Rate: Total number of Data packets
successfully retrieved to interest packets in the net-
work.
6) Query Satisfaction Rate: Ratio of total successfully
satisfied data against a total of query-based interest
sent.
7) Energy Consumption: The aggregate energy is utilized
by the nodes for the transmission of the interest packet.
5 10 15 20
0
50
100
150
200
INF:NDN IoT
SICN
Length (Characters)
Number of names (Interest/Data)
FIGURE 17: Content name length
B. EXPERIMENTAL RESULTS
1) Name Length
The average name length as a function of a number of content
names (interest and data) can be visualized in Fig. 17. Results
demonstrate that the average length of content names is
decreased by approximately 40.05 % compared to SICN. The
rationale is that SICN does not incorporate any mechanisms
to shorten the length of content names. In contrast, INF-NDN
IoT effectively reduces the character count of content names
by leveraging various NLP techniques such as data cleaning,
normalization, abbreviation, stemming, stopword removal,
and segmentation.
Our findings demonstrate the effectiveness of incorporat-
ing NLP techniques in streamlining content names, resulting
in more concise names. It is to be noted that name lengths
are related to entries of content names in CDB-based FIB.
The name length does not include the semantic tag. The
rationale behind this is that INF-NDN IoT removes the tag
while creating an entry in the CDB-based FIB.
2) Name Memory Utilization
Memory utilization is directly proportional to the length of
names. A larger length of NDN name increases the memory
footprints. Fig. 18 demonstrates the relationship between
memory utilization (in megabytes) of content names and the
total number of content names with respect to INF-NDN
IoT and SICN. Results show that INF-NDN IoT improves
memory utilization by harnessing the capabilities of NLP,
resulting in shorter content names. INF-NDN IoT does not
introduce extra components or information in the structure
of the NDN name. On the contrary, SICN incorporates ad-
ditional identifiers, increasing its overall length and memory
footprint. In addition, SICN also includes longer and more
descriptive names and repetitions of “/” or “$”. These extra
characters increase the overall memory utilization.
12345678910
0
20
40
60
80
100
120
140
160
180
200
Memory utlization (MB)
Number of names (Million)
INF: NDN IoT
SICN
FIGURE 18: Content name memory
VOLUME 4, 2016 15
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
20 25 30 35 40 45
2
4
6
8
10
12
14
16
18
Average hop count
Number of nodes
INF: NDN IoT
SICN
FIGURE 19: Average number of hop counts
5 10 15 20 25
0
2
4
6
8
10
12
14
16
Energy consumption (Jouls)
Number of Interest
INF: NDN IoT
SICN
FIGURE 20: Interest energy consumption
3) Number of Hops
Fig. 19 shows the average no of hops count on the y-axis
as a function of the number of nodes on the x-axis. Results
demonstrate that the INF-NDN IoT reduces the number of
hops required to reach the destination. One core reason for
the reduction in the number of hops is that all supernodes are
assigned semantic tags and share CDB. Due to the semantic
tags, supernodes perform semantic and selective forwarding.
Semantic-based forwarding enables supernodes to select in-
terfaces aligning with the specific semantic requirements of
the transmitted data.
As a result, not all nodes participate in packet forwarding.
By involving fewer nodes in the forwarding process, the
number of hops a packet needs to traverse is 4 when overall
network nodes are 16. If numbers of nodes are increased
to 50, the average hop count for INF-NDN IoT is 10. On
the contrary, SICN’s average hop count is 7 and 15 when
a total number of nodes is 16 and 50 respectively, which is
more than INF-NDN IoT. SICN forwards data on behalf of
matching IDs, namely, geo, semantic, and publisher IDs.
In the scenario of mismatching IDs or mobility of pub-
lisher and consumer, SICN performs broadcasting. As a
5 10 15 20 25
0
20
40
60
80
100
Interest satisfaction rate (%)
Number of Interest
INF: NDN IoT
SICN
FIGURE 21: Interest satisfaction rate
5 10 15 20 25
20
40
60
80
100
Query satisfaction rate
Query based Interest
INF: NDN IoT
SICN
FIGURE 22: Query satisfaction rate
result, the average hop count is increased.
4) Energy Consumption
Total energy consumption as a function of the number of
interest packets is illustrated in Fig. 20. INF-NDN IoT
consumes less energy than SICN in the case of forwarding
interest packets. The reason is that the principal node for-
wards the interest packet to the producer supernode. Same
as, all ordinary nodes directly communicate with supernodes,
eliminating the need for any intermediate node. In the worst
case, if the supernode doesn’t have data, supernodes directly
forward the interest packet to the producer supernode due
to CDB. This also minimizes the intermediate nodes. As a
result, INF-NDN IoT prevents the occurrence of broadcast
storms and minimizes collisions in the network. Conse-
quently, the overall energy consumption across the network
is significantly reduced. In contrast, SICN uses a hybrid
approach. It efficiently handles interest flooding by matching
three-dimensional IDs. If there’s a match, forwarding occurs
smoothly; otherwise, SICN broadcasts the interest when IDs
don’t match. As a result, the energy consumption of SICN is
higher than INF-NDN IoT.
16 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
5) Interest Satisfaction Rate
Fig. 21 shows the relationship between interest packets and
ISR. we increase interest packets from 0 to 25 per second.
Results show that INF-NDN IoT performs better. The reason
is that INF-NDN IoT directly fetches the content from the
producer. The principal node directly forwards the interest
packet to the relevant producer or supernode and fetches
the data without involving intermediate nodes. This semantic
awareness leads to a higher interest satisfaction rate as it
improves the precision and accuracy of data retrieval. On the
contrary, SICN does not explicitly incorporate semantic tags
in the NDN name. It primarily focuses on publisher ID and
geographical information and broadcasts in the worst case,
thus decreasing the interest satisfaction rate. SICN schemes
involve multiple intermediate forwarders in the communica-
tion, causing congestion and collisions in the network and
thus decreasing ISR.
6) Query Satisfaction Rate
Fig. 22 illustrates the query satisfaction rate against query-
based interests. Results demonstrate that INF-NDN IoT
achieves a higher satisfaction rate as compared to SICN. The
reason is a correlation between semantic tags and supernodes.
The correlation between tags and supernodes has categorized
the smart campus network into distinct contextual smart ar-
eas, each of which is identified by its corresponding tags. The
contextual nature of the query-based interest name is used by
the principal node in forwarding the interest to the supernode
that shares the same semantic tag. As a result, semantic data
retrieval is achieved which is directly proportional to the
query satisfaction rate. On the contrary, SICN only supports
query satisfaction rate when semantic ID is matched either
with geo ID or publisher ID in the forwarding table. Other-
wise, SICN does not provide semantic data retrieval if there
are non-semantic IDs to match.
7) Interest Retrieval Time
Interest retrieval time (IRT) results of INF-NDN IoT and
SICN are illustrated in Fig. 23. We increase the number of
interest packets from 0 to 25 per second. Results clearly
demonstrate that INF-NDN IoT shows a substantial reduction
in retrieval time. The reason is that INF-NDN IoT develops
query optimization mechanisms enriched with semantic tags.
Optimized query directly goes to the relevant supernode with
the same semantic tag and fetches precise and most relevant
data. In this way, the delay associated with retrieving the
identical data packet is significantly reduced compared to
sending multiple interest requests. This obviates the need
for multiple, separate interest requests for each individual
data packet. Overall, INF-NDN IoT outperforms the SICN
with 78 % to 90 % in retrieval time. On the contrary, SICN
consumes additional time on every node in the calculation
of matching ID. In SICN, the geographic IDs of publishers
and consumers are changed with their physical mobility (not
network or channel mobility).
5 10 15 20 25
0
100
200
300
400
500
600
700
Interest retrieval time (ms)
Number of Interest
INF: NDN IOT
SICN
FIGURE 23: Interest retrieval time
In addition to exploring the scalability of INF-NDN IoT,
we choose IRT as a scalability metric. We increased the
number of interest packets in increments of 50. We observed
that with each increment of 50 interest packets per second,
the IRT increased by an average of 0.06 second. Therefore,
it is important to note that our solution does not introduce
significant overhead with the increasing number of interest
packets, as measured by IRT.
The rationale for selecting IRT as a scalability metric is
that the proposed scheme is specifically designed for small
IoT environments. INF-NDN IoT is well-suited for networks
where traffic can be categorized into distinct semantic areas,
a capability that is not feasible in very large-scale networks
such as [52, 53].
VII. CONCLUSION AND FUTURE WORK
In this article, we proposed INF-NDN IoT for intelligent
naming and forwarding. To accomplish this, INF-NDN IoT
applies NLP to optimize the content names and remove name
ambiguities. Afterward, semantic information is obtained
from content names, and semantic tags are assigned. In
parallel, INF-NDN IoT finds the set of backbone nodes in the
network called supernodes, to support semantic-assisted for-
warding. These supernodes are also assigned semantic tags
to develop a correlation between naming and forwarding. To
demonstrate the proposed methodology’s effectiveness, the
principal node is responsible for the selection of supernodes
and NLP-based functionalities.
As demonstrated by the results, the principal node effec-
tively resolves any name ambiguity in the content name and
forwards the interest packet based on the semantics of the
content names. In addition, INF-NDN IoT reduces the aver-
age content name’s length up to approximately 40.05% and
memory utilization by removing unnecessary characters from
content names. INF-NDN IoT surpasses the target scheme
with 70 to 90 % in retrieval time. Moreover, the proposed
scheme also achieves a higher interest satisfaction rate and
lower energy consumption, thus outperforming the state-of-
VOLUME 4, 2016 17
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
the-art benchmark.
The architecture of INF-NDN IoT meets the essential
needs for the basic requisites of NDN IoT. However, it
marks just the initial stage of developing a concrete NDN-
based architecture for IoT. Many important aspects remain
unexplored in this scheme and will be addressed in our future
work. One of these challenges is the scalability in terms
of infrastructure. The scope of INF-NDN IoT is currently
limited to small-scale environments, where ICN network
traffic can be categorized into distinct semantic groups. This
semantical categorization also presents scalability challenges
in maintaining smooth and frequent communication across
interdomain networks. Currently, only the principal node
handles naming tasks and assigns semantic tags. Introduc-
ing hundreds of supernodes also requires hundreds of se-
mantic tags in the network. In addition, it is not feasible
to semantically categorize all nodes in a very large-scale
network. Therefore, the implementation of INF-NDN IoT
on a very large-scale network which includes thousands of
network nodes is yet to be resolved. To address this, we
plan to develop a universal semantic naming and forwarding
scheme designed to facilitate effective communication across
domains. We also intend to create a decentralized intelligent
method in the future that can operate independently at the
local level. Moreover, the semantic evaluation of CS and PIT
also has been deferred to future work.
REFERENCES
[1] S. Li, L. D. Xu, S. Zhao, The internet of things: a survey,
Information systems frontiers 17 (2015) 243–259.
[2] K. Abualsaud, T. M. Elfouly, T. Khattab, E. Yaacoub,
L. S. Ismail, M. H. Ahmed, M. Guizani, A survey on
mobile crowd-sensing and its applications in the iot era,
IEEE Access 7 (2018) 3855–3881.
[3] H. Xu, Y. He, X. Sun, J. He, Q. Xu, Prediction of ther-
mal energy inside smart homes using iot and classifier
ensemble techniques, Computer Communications 151
(2020) 581–589.
[4] Q. Cai, H. Wang, Z. Li, X. Liu, A survey on multimodal
data-driven smart healthcare systems: approaches and
applications, IEEE Access 7 (2019) 133583–133599.
[5] Z. Zhang, E. Lu, Y. Li, L. Zhang, T. Yu, D. Pesavento,
J. Shi, L. Benmohamed, Ndnot: a framework for named
data network of things, in: Proceedings of the 5th ACM
Conference on Information-Centric Networking, 2018,
pp. 200–201.
[6] G. Mauri, M. Gerla, F. Bruno, M. Cesana, G. Verti-
cale, Optimal content prefetching in ndn vehicle-to-
infrastructure scenario, IEEE Transactions on Vehicular
Technology 66 (3) (2016) 2513–2525.
[7] A. Nurhayati, R. Mayasari, S. Ahdan, R. M. Negara,
G. N. Nurkahfi, N. R. Syambas, Naming scheme on
named data networking: A survey, in: 2022 8th Interna-
tional Conference on Wireless and Telematics (ICWT),
IEEE, 2022, pp. 1–5.
[8] Z. Sabir, A. Amine, Connected vehicles using ndn
for intelligent transportation systems, in: International
Conference on Industrial Engineering and Operations
Management, Vol. 2018, 2018, pp. 2433–2441.
[9] Z. Li, J. Liu, L. Yan, B. Zhang, P. Luo, K. Liu, Smart
name lookup for ndn forwarding plane via neural net-
works, IEEE/ACM Transactions on Networking 30 (2)
(2021) 529–541.
[10] S. Shannigrahi, C. Fan, C. Partridge, What’s in a
name? naming big science data in named data network-
ing, in: Proceedings of the 7th ACM Conference on
Information-Centric Networking, 2020, pp. 12–23.
[11] F. A. Karim, A. H. M. Aman, R. Hassan, K. Nisar,
M. Uddin, Named data networking: A survey on routing
strategies, IEEE Access 10 (2022) 90254–90270.
[12] D. Khurana, A. Koli, K. Khatter, S. Singh, Natural
language processing: State of the art, current trends and
challenges, Multimedia tools and applications 82 (3)
(2023) 3713–3744.
[13] G. M. Raza, Z. S. Butt, S. Latif, A. Wahid, Sentiment
analysis on covid tweets: an experimental analysis on
the impact of count vectorizer and tf-idf on sentiment
predictions using deep learning models, in: 2021 Inter-
national Conference on Digital Futures and Transfor-
mative Technologies (ICoDT2), IEEE, 2021, pp. 1–6.
[14] T. Song, H. Yuan, P. Crowley, B. Zhang, Scalable
name-based packet forwarding: From millions to bil-
lions, in: Proceedings of the 2nd ACM conference on
information-centric networking, 2015, pp. 19–28.
[15] B. S. Paris, Time constructs: Design ideology and a
future internet, Time & Society 30 (1) (2021) 126–149.
[16] A. Detti, L. Bracciale, P. Loreti, G. Rossi, N. B.
Melazzi, A cluster-based scalable router for information
centric networks, Computer networks 142 (2018) 24–
32.
[17] C. Gündo˘
gan, C. Amsüss, T. C. Schmidt, M. Wählisch,
Content object security in the internet of things: Chal-
lenges, prospects, and emerging solutions, IEEE Trans-
actions on Network and Service Management 19 (1)
(2021) 538–553.
[18] A. Aboodi, T.-C. Wan, G.-C. Sodhy, Survey on the
incorporation of ndn/ccn in iot, IEEE Access 7 (2019)
71827–71858.
[19] X. Tan, W. Feng, J. Lv, Y. Jin, Z. Zhao, J. Yang, F-ndn:
An extended architecture of ndn supporting flow trans-
mission mode, IEEE Transactions on Communications
68 (10) (2020) 6359–6373.
[20] R. Paul, M. P. Selvan, A study on naming and caching
in named data networking, in: 2021 fifth international
conference on I-SMAC (IoT in social, mobile, analytics
and cloud)(I-SMAC), IEEE, 2021, pp. 1387–1395.
[21] K. Ahed, M. Benamar, R. El Ouazzani, Content deliv-
ery in named data networking based internet of things,
in: 2019 15th International Wireless Communications
& Mobile Computing Conference (IWCMC), IEEE,
2019, pp. 1397–1402.
[22] H. Hellbrück, T. Teubler, S. Fischer, Name-centric
18 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
service architecture for cyber-physical systems (short
paper), in: 2013 IEEE 6th International Conference on
Service-Oriented Computing and Applications, IEEE,
2013, pp. 77–82.
[23] C. Seales, T. Do, E. Belyi, S. Kumar, Phinet: A plug-
n-play content-centric testbed framework for health-
internet of things, in: 2015 IEEE International Confer-
ence on Mobile Services, IEEE, 2015, pp. 368–375.
[24] P. Hemanth, Home automation using resberrypi &
cloud, Global Journal of Engineering Science and Re-
search Management 6 (2016) 42–48.
[25] M. I. Ayadi, A. Maizate, M. Ouzzif, C. Mahmoudi,
Deep learning in building management systems over
ndn: Use case of forwarding and hvac control, in:
2019 International Conference on Internet of Things
(iThings) and IEEE Green Computing and Commu-
nications (GreenCom) and IEEE Cyber, Physical and
Social Computing (CPSCom) and IEEE Smart Data
(SmartData), IEEE, 2019, pp. 1192–1198.
[26] D. Mars, S. Mettali Gammar, A. Lahmadi,
L. Azouz Saidane, Using information centric
networking in internet of things: a survey, Wireless
Personal Communications 105 (2019) 87–103.
[27] T. Niu, F. Yang, An entropy-based approach: Com-
pressing names for ndn lookup, IEEE Access 9 (2021)
109833–109846.
[28] A. Javaheri, S. N. S. Hashemi, A. Bohlooli, Hybrid
naming scheme based pursuit architecture for smart
city, in: 2020 4th International Conference on Smart
City, Internet of Things and Applications (SCIOT),
IEEE, 2020, pp. 33–38.
[29] B. Nour, K. Sharif, F. Li, H. Moungla, Y. Liu, M2hav:
A standardized icn naming scheme for wireless de-
vices in internet of things, in: Wireless Algorithms,
Systems, and Applications: 12th International Confer-
ence, WASA 2017, Guilin, China, June 19-21, 2017,
Proceedings 12, Springer, 2017, pp. 289–301.
[30] M. A. U. Rehman, R. Ullah, B. S. Kim, Ninq: Name-
integrated query framework for named-data networking
of things, Sensors 19 (13) (2019) 2906.
[31] L. Bracciale, P. Loreti, A. Detti, R. Paolillo, N. B.
Melazzi, Lightweight named object: An icn-based ab-
straction for iot device programming and management,
IEEE Internet of Things Journal 6 (3) (2019) 5029–
5039.
[32] G. Jaber, N. Pastei, F. Rahal, A. Abboud, Naming and
routing scheme for data content objects in information-
centric network, in: 2020 IEEE Open Conference of
Electrical, Electronic and Information Sciences (eS-
tream), IEEE, 2020, pp. 1–5.
[33] T. Mochida, D. Nozaki, K. Okamoto, X. Qi, Z. Wen,
T. Sato, K. Yu, Naming scheme using nlp machine
learning method for network weather monitoring sys-
tem based on icn, in: 2017 20th international sympo-
sium on wireless personal multimedia communications
(WPMC), IEEE, 2017, pp. 428–434.
[34] X. Wang, S. Cai, Edge-assisted ndn-based iot frame-
work with provider and consumer mobility support,
IEEE Transactions on Network Science and Engineer-
ing 9 (3) (2022) 1713–1725.
[35] F. Song, Y. Ma, Z. Yuan, I. You, G. Pau, H. Zhang,
Exploring reliable decentralized networks with smart
collaborative theory, IEEE Communications Magazine
(2023).
[36] T. Chatterjee, S. Ruj, S. Das Bit, Efficient data storage
and name look-up in named data networking using
connected dominating set and patricia trie, Automatic
Control and Computer Sciences 55 (2021) 319–333.
[37] F. KARIM, K. A. Abu Bakar, N. H. Kamarudin, et al.,
Explicit routing for popular content in named data
network, Azana Hafizah and Abu Bakar, Khairul Azmi
and Kamarudin, Nazhatul Hafizah, Explicit Routing for
Popular Content in Named Data Network.
[38] F. Wang, H. Li, Z. Wu, P. Lu, B. Wang, Y. Yang, H. Bai,
X. Pei, Alrp: An adaptive link state routing protocol
for multi-identifier network, in: 2022 IEEE 8th Interna-
tional Conference on Computer and Communications
(ICCC), IEEE, 2022, pp. 677–684.
[39] M. A. U. Rehman, R. Ullah, B.-S. Kim, B. Nour,
S. Mastorakis, Ccic-wsn: An architecture for single-
channel cluster-based information-centric wireless sen-
sor networks, IEEE Internet of Things Journal 8 (9)
(2020) 7661–7675.
[40] A. Aboud, H. Touati, B. Hnich, Efficient forwarding
strategy in a ndn-based internet of things, Cluster Com-
puting 22 (2019) 805–818.
[41] A. Djama, B. Djamaa, M. R. Senouci, N. Khemache,
Lafs: a learning-based adaptive forwarding strategy for
ndn-based iot networks, Annals of Telecommunications
77 (5-6) (2022) 311–330.
[42] N. Akther, K. Dhar, S. M. A. Iqbal, M. N. Huda, et al.,
Interest forwarding strategy in named data networks
(ndn) using thompson sampling, Journal of Network
and Computer Applications 205 (2022) 103458.
[43] E. da Silva, J. Macedo, A. Costa, Context-aware routing
and forwarding model for ndn-based vanet, in: Inter-
national Conference on Ad Hoc Networks, Springer,
2021, pp. 18–32.
[44] X. Wang, Y. Lu, Efficient forwarding and data acquisi-
tion in ndn-based manet, IEEE Transactions on Mobile
Computing 21 (2) (2020) 530–539.
[45] D. M. Blei, J. D. Lafferty, Topic models, in: Text Min-
ing, 1st Edition, Chapman and Hall/CRC, 2009, p. 24.
[46] D. M. Blei, A. Y. Ng, M. I. Jordan, Latent dirichlet al-
location, Journal of machine Learning research 3 (Jan)
(2003) 993–1022.
[47] G. Veletsianos, R. Kimmons, R. Larsen, T. A. Dousay,
P. R. Lowenthal, Public comment sentiment on educa-
tional videos: Understanding the effects of presenter
gender, video format, threading, and moderation on
youtube ted talk comments, PloS one 13 (6) (2018)
e0197331.
VOLUME 4, 2016 19
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
Author et al.: Preparation of Papers for IEEE Access
[48] J. Li, J. Wu, C. Li, W. Yang, A. K. Bashir, J. Li, Y. D. Al-
Otaibi, Information-centric wireless sensor networking
scheme with water-depth-awareness content caching
for underwater iot, IEEE Internet of Things Journal
9 (2) (2021) 858–867.
[49] Z. Kong, E. Newberry, J. Shi, T. Yu, A. Kahlert,
D. Hensley, Python-ndn, https://python-ndn.
readthedocs.io/en/latest/, accessed: 2023 (Year).
[50] N. P. Team, et al., Pyndn2.(2017) (2017).
[51] N. Hardeniya, J. Perkins, D. Chopra, N. Joshi,
I. Mathur, Natural language processing: python and
NLTK, Packt Publishing Ltd, 2016.
[52] H. Touati, A. Aboud, B. Hnich, Named data
networking-based communication model for internet
of things using energy aware forwarding strategy and
smart sleep mode, Concurrency and Computation: Prac-
tice and Experience 34 (3) (2022) e6584.
[53] F. Azzedin, Towards a scalable hdfs architecture, in:
2013 international conference on collaboration tech-
nologies and systems (CTS), IEEE, 2013, pp. 155–161.
GHULAM MUSA RAZA received his BS degree
in Computer Sciences from Comsats University
Islamabad in 2019. His major in BS was Intel-
ligent Robotics. He received his MS degree in
Computer Sciences from SEECS, NUST Islam-
abad in 2021. His research interest in his Master
was Natural Language Processing (Artificial In-
telligence). From 2017 to 2019, he was working
as a Software Engineer at Snaky Solutions Pvt
Limited. He served as a Machine Learning based
Research Assistant in the TUKL lab, NUST Islamabad at the start of 2021.
He served as Lecturer at Alhamd Islamic University, Islamabad from 2021 to
2022. His major interests are in the field of Natural Language Processing, the
Internet of Things (IOT), Information-Centric Networking, and Named Data
Networking. He is currently pursuing a Ph.D. degree with the Department
of Communication and Software Engineering at Graduate School, Hongik
University, South Korea.
IHSAN ULLAH received his B.S. degree in Com-
puter Systems engineering from the University
of Engineering and Technology Peshawar, Pak-
istan, and his Master degree in computer engi-
neering with a specialty in computer and Wire-
less Networks from the Department of Electrical
and Computer Engineering, COMSATS Univer-
sity, Islamabad, Pakistan in 2021. He worked as
a research assistant in the Wireless and Commu-
nication lab for six months. Currently, he is doing
his Ph.D. in the Department of Software and Communication Engineering,
at Hongik University, South Korea. His current interests are in the area of
NDN, Underwater Wireless Sensor Networks (UWSN), Cloud computing,
Fog Computing, Vehicular Networks, Machine learning, and Artificial intel-
ligence.
MUHAMMAD SALAH UD DIN is currently pur-
suing a Ph.D. degree in Computer Engineering
with the Broadband Convergence Networks Lab-
oratory at Hongik University, South Korea. He
received a B.E. degree in Information & Technol-
ogy from The KICSIT (University of Engineering
and Technology, Taxila) Pakistan, in 2013, and
an M.S. degree in Computer Science from COM-
SATS University, Islamabad, Pakistan, in 2016.
From 2013 to 2019, he worked as a Software
Engineer in leading IT companies in Pakistan. His major interests are in
the field of Information-Centric Wireless Networks, Vehicular networks,
Named Data Networking, Wireless Sensor Networking, Edge/Fog/Cloud
Computing, and AI for Wireless communications.
MUHAMMAD ATIF UR REHMAN is an Assis-
tant Professor at Manchester Metropolitan Uni-
versity, Manchester, UK. He received his PhD
degree in Electronics and Computer Engineering
from Hongik University, South Korea in 2022. His
research focuses on distributed networking and
computing, Metaverse, edge computing, IoT, UAV,
ICN/NDN, and 6G. He is a founding member and
a part of the IEEE SIG on Metaverse. He served as
a General Chair, TPC member, and Session Chair
for several conferences and workshops such as IEEE ICC, iMeta, MetaNC,
and ICWMC, among many others. His work has appeared in more than 50
publications in reputable venues such as IEEE Internet of Things Journal,
ACM SIGCOMM, IEEE Intelligent Transporation System Magazine, and
Future Generation Computer Systems, to name a few. He currently holds
six patents. In addition to his academic background, he also has Industry
experience of over 6 years as a Software Engineer & Architect in leading
companies.
BYUNG-SEO KIM received his B.S. degree in
Electrical Engineering from In-Ha University, In-
Chon, Korea in 1998 and his M.S. and Ph.D.
degrees in Electrical and Computer Engineering
from the University of Florida in 2001 and 2004,
respectively. His Ph.D. study was supervised by
Dr. Yuguang Fang. Between 1997 and 1999, he
worked for Motorola Korea Ltd., PaJu, Korea
as a Computer Integrated Manufacturing (CIM)
Engineer in Advanced Technology Research and
Development (ATR & D). From January 2005 to August 2007, he worked
for Motorola Inc., Schaumburg Illinois, as a Senior Software Engineer in
Networks and Enterprises. His research focuses on Motorola Inc, design-
ing the protocol and network architecture of wireless broadband mission-
critical communications. From 2012 to 2014, he was the Chairman of
the Department of Software and Communications Engineering, at Hongik
University, South Korea, where he is currently a Professor. He served as the
General Chair for General Chair of 3rd IWWCN 2017, and the TPC member
for the IEEE VTC 2014-Spring and the EAI FUTURE2016, and ICGHIC
2016 2019 conferences. He served as Guest Editor of special issues of the
International Journal of Distributed Sensor Networks (SAGE), IEEE Access,
and the Journal of the Institute of Electrics and Information Engineers. He
also served as a Member of the Sejong-city Construction Review Committee
and the Ansan-city Design Advisory Board. His work has appeared in
around 167 publications and 22 patents. He is an IEEE Senior Member
and Associative Editor of IEEE Access. His research interests include the
design and development of efficient wireless/wired networks including link-
adaptable/ cross-layer-based protocols, multi-protocol structures, wireless
CCNs/NDNs, Mobile Edge Computing, physical layer design for broadband
PLC, and resource allocation algorithms for wireless networks.
20 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3444903
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
