Sie sind auf Seite 1von 9

IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO.

2, APRIL 2018 491

ISI: Integrate Sensor Networks


to Internet With ICN
Sripriya Srikant Adhatarao, Mayutan Arumaithurai, Dirk Kutscher, and Xiaoming Fu

Abstract—Internet of Things (IoT) is a growing topic of applications in many different fields like smart homes, smart
interest. Billions of IoT devices are expected to connect to the vehicles, industries, environmental monitoring, etc.
Internet in the near future. These devices differ from the tradi- Billions of IoT devices are expected to be connected in the
tional devices operated in the Internet. In this paper, we argue
that an information-centric networking (ICN), a new networking near future to communicate, sense and gather data. However, it
paradigm, is a more suitable architecture for the IoT compared is important to understand the difference between IoT and sen-
to the currently prevailing IP-based network. We observe that sor networks. There are various heterogenous sensor networks
recent works that propose to use ICN for IoT, either do not cover that operate in their own private network. The goal of IoT is
the need to integrate sensor networks with the Internet to realize to provide the sensor networks with access to Internet in order
IoT or do so inefficiently. There is a need to understand effective
ways to integrate the various heterogeneous sensor networks with for them to become IoT [1].
the Internet without affecting their current mode of operation. Currently, IoTs are designed to operate with the IP architec-
In this paper, we study the essential requirements for integrating ture [2]. However, IoT networks often contain many resource
sensor networks to the Internet. We provide an architecture with constrained devices with smaller memory, limited computa-
gateways for paving a way for the sensor networks to become a tional capacity, and power supply (mostly a battery). Many
part of the IoT family. We further provide a naming schema for
efficient operation of the resource constrained sensor networks, IoT applications require devices to operate for longer periods
discuss mobility, security, communication patterns, and propose in remote locations with no facilities, e.g., forests. Due to con-
the most suitable choices for IoT networks. straints, IoT devices are equipped with Layer2 technologies
Index Terms—Gateway, information-centric networking (ICN), like IEEE 802.15.4 and Bluetooth LE; hence, they operate
Internet, Internet of Things (IoT), sensor networks. with a much smaller MTU than the current MTU used in the
Internet. They also incur several other challenges like limited
IP address space, while point-to-point connectivity is heavy
for these resource constrained devices. Additionally security
I. I NTRODUCTION is another critical aspect in many IoT applications and is
NTERNET of Things (IoT) refers to a network of devices expensive (induces overhead) to achieve with IP leading to
I like machines, vehicles, electronic appliances, and also
wearables like radio frequency identification tags, step-counter,
complexity in operation and resource consumption. There is
therefore a need for an efficient design for the IoT devices
etc. These devices are usually embedded with sensors, actu- that is scalable, efficient, and provides a secure mechanism
ators, memory, and network connectivity. They are mainly for communication to gather data for monitoring and/or con-
used for sensing, monitoring, and controlling various appli- trolling the devices. Shang et al. [3] discussed many of such
cations. IoT is a growing topic of interest and has already issues in detail.
drawn attention of academia, and industry. We observed that IoTs are usually studied/researched as sep-
A current popular topic of interest in the IoT commu- arate entities (see Section X). However, that should not be the
nity is smart cities. The smart cities are in design phase case. One has to also consider all the potential issues in the
and will come to reality in the near future. The IoT devices Internet. Isolating them might lead to unforseen consequences.
will be deployed extensively in the infrastructures like smart IoT needs well connected networks to communicate and pass
buildings, smart offices, etc. Additionally, IoT’s have found information/control messages to other devices in the network.
One might also question, what is the benefit of integrating the
Manuscript received December 28, 2016; revised July 24, 2017; accepted sensor networks with the Internet? We believe that by integrat-
August 11, 2017. Date of publication August 18, 2017; date of cur- ing the sensor networks with the Internet, both the networks
rent version April 10, 2018. This work was supported by the joint can benefit. Primarily, sensor networks will benefit from the
EU H2020/NICT ICN2020 Project under Contract 723014 and Contract
NICT 184. (Corresponding author: Mayutan Arumaithurai.) existing features of the Internet thus incorporating the ubiq-
S. S. Adhatarao and M. Arumaithurai are with the Institut für uitous sensor networks into the IoT world. At the same time,
Informatik, University of Göttingen, 37077 Göttingen, Germany (e-mail: we see a scope for the Internet to widen.
adhatarao@cs.uni-goettingen.de; arumaithurai@cs.uni-goettingen.de).
D. Kutscher is with Huawei’s German Research Center, 80992 Munich, Information-centric networking (ICN) [4], [5] is a new
Germany (e-mail: dku@dkutscher.net). networking paradigm. In ICN, content is treated as the first-
X. Fu is with the Computer Networks Group, Institut für class entity and nodes exchange information based on the
Informatik, University of Göttingen, 37077 Göttingen, Germany (e-mail:
fu@cs.uni-goettingen.de). Names of the content instead of the IP addresses of the end
Digital Object Identifier 10.1109/JIOT.2017.2741923 points that request or provide the information. This shift from
2327-4662 c 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
492 IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 2, APRIL 2018

a “location-based” network to a “content-centric” network


entails efficiency for content dissemination, especially when
the content may be available at multiple points and also when
the provider and/or consumer are mobile. Since many ICN
designs incorporate extensive (in-network) caching, additional (a) (b)
performance benefits can be realized with those widespread
caches. ICN is growing rapidly with a highly active commu-
nity. Researchers are increasingly proposing new and improved
solutions in various areas like routing, forwarding, caching,
naming, etc.
Similar to ICN, in IoT the devices are interested in the con-
tent and not their location, i.e., IoT’s are information centric
(c) (d)
in nature. The design of ICN suits fairly well for IoTs. There
is no longer a need for maintaining point-point communica- Fig. 1. Use cases. (a) Scenario1. (b) Scenario2. (c) Scenario3. (d) Scenario4.
tion. Recent ICN proposals, such as named data networking
(NDN) [4] and content oriented publish/subscribe (Pub/Sub)
system (COPSS) [5], which enhances NDN with an effi-
this vital need. The main goal of this paper is to connect the
cient Pub/Sub capability, adopt human-readable, hierarchically
multitude of sensor networks to the Internet, i.e., to connect
structured names and content descriptors (CDs). The names-
these two worlds together. We look at the various require-
pace is unbounded and can easily support billions of devices
ments for realizing such an ICN-based – IoT = Internet +
and even more. There is an increased availability of data due
SN architecture. We propose to introduce gateways to inte-
to in-network caching. Security is enhanced since ICN embeds
grate the sensor networks with the Internet. We identify the
security in the content rather than on the communication
explicit functions/responsibilities of such a gateway and the
link. The multicast and broadcast features are easily supported
various services offered by it. We design a naming schema that
with ICN without additional overhead. Although, there have
can efficiently support the sensor networks running ICN proto-
been many instances, where NDN has been explored for the
cols. Additionally, we also analyze the various communication
Internet and in some cases for supporting the IoT environ-
patterns, mobility, and security for IoT devices.
ments [3], [6], [7] we argue that, IoT does not need the full
The contributions in this paper include the following.
content centric networking (CCN/NDN) stack. Some other
1) Analysis of the requirements for an architecture to
works like CCN-lite/NDN-lite [8] support more specifically
integrate sensor networks with the Internet.
the resource constrained IoT devices.
2) Architecture ISI: To integrate sensor networks to the
The Internet can represent either the existing TCP/IP, ICN
Internet with ICN.
or both the architectures. In this paper, we mainly focus on
3) Naming schema for the IoT networks running ICN.
the Internet running ICN protocols because the various Internet
4) Communication protocol for the IoT networks.
Engineering Task Force (IETF) groups (e.g., 6LoWPAN [9],
5) Discussion on the aspect of mobility and security
Core, and ROLL) [10] are already focusing on integrating the
for IoT.
IoT devices to the IP-based Internet. With ICN in the Internet,
we see an increased amount of caching in the Internet. Hence,
the IoT networks will incur less traffic. On the other hand,
the Internet users can have access to IoT data without having II. U SE C ASE F ORMALIZATION
to explicitly follow a different protocol for the multitude of In this section we formalize some of the essential use
IoT networks. The users can also control the IoT devices via cases that will derive the requirements for integrating sensor
Internet. This can also simplify the application designs. networks to the Internet and will further assist in building the
Many recent works that advocate the use of ICN for ISI architecture.
IoT focus on aspects, such as data retrieval patterns [11], We take the example of a smart city as shown in Fig. 1
routing [12], benefits of caching [13], and architectural with many buildings equipped with sensor networks to build
changes [14]. The Internet houses powerful devices that are our use cases. Let us consider an application of monitoring
capable of running the full fledged ICN protocols like NDN. the temperature of all the rooms in a smart building as a
However, the sensor networks do not need the entire ICN representative sensor network. The smart building has sev-
stack and would require lighter versions of the ICN protocols. eral rooms equipped with temperature sensitive machinery,
Hence, like recent works we assume that the sensor networks e.g., servers and hence each room is equipped with a tem-
will operate with lighter versions of the ICN protocols. perature sensing sensor device. There is a base station (BS)
Even though the recent works like [12], [14], etc. have that gathers the temperature from each room every 30 min
focused on solving many different problems of IoT, they have and raises an alarm when abnormalities occur. Several such
not considered the crucial aspect of the need for an architec- scenarios can be gathered in a smart city, e.g., in industrial
ture to integrate the sensor networks running lighter version of units, offices, smart houses for controlling fire, etc. Let us
ICN protocol with the Internet in order to pave a way for them say, the sensor network1 (SN1) is operating in the building1
to join the IoT family. Therefore, in this paper, we address which is a computer science building in a university, and
ADHATARAO et al.: ISI WITH ICN 493

sensor network2 (SN2) is operating in the building2, which A. Gateway


is a mathematics building in the same university. We observe from use cases 1–3 that there are multitude
Scenario 1: A user (U1) (e.g., a system administrator) in of users scattered across hybrid sensor networks and Internet.
the Internet is interested in the temperature of room1 in the However, these networks do not use the same protocol. There
building1. U1 is in Internet that runs an ICN protocol like is a need for protocol translation among these networks for
NDN [15]. However, the building1 is located in SN1 that runs communicating with each other. An efficient way to interface
a lighter version of an ICN protocol like CCN-lite [8]. Besides, networks operating different protocols is via a gateway. The
U1 is interested in the temperature of room1 sensed by the gateway should run both the protocols of the networks it
sensor device S1 which is not awake all the time in order to serves. The gateway should be equipped with necessary intel-
save its constrained resources like power. U1 is not aware of ligence and data structures to perform near transparent flow of
the time at which S1 will be available to serve the requested traffic between the two networks to seamlessly integrate these
content, nor the exact protocol that it runs. networks.
Scenario 2: A user (U2) in the Internet is interested in
receiving the cumulative temperature of the building1 every
30 mins. Similar to scenario1, U2 is in Internet but would B. Naming
like to receive the content collected and computed by the BS Since ICN is a name-based protocol, we observe from
located in the sensor network. all the four use cases a need for a naming schema for the
Scenario 3: The BS1 in SN1 needs some content, e.g., the IoT devices that is more catered toward the operation of
GPS location of the mathematics building in the smart city. the constrained devices in the sensor networks. The rationale
The mathematics building is located in SN2 of the smart city. for this requirement is the MTU in the sensor networks is
The SN2 might be running the same or a different ICN proto- much smaller (127B) when compared to the MTU (1280B)
col from SN1. BS1 is un aware of the location of the content used by the devices in the Internet. ICN supports, unbounded
or the protocol used in SN2. and any length namespace. There is a need for smaller ICN
Scenario 4: Let us consider the operation of SN1. The sen- names that can not only fit into the MTU of the sensor
sor devices sense the temperature in each room periodically networks devices but also requires less storage on the forward-
every 30 min. The BS1 is interested in the content produced ing engines of the sensing devices. This will ensure that IoT
by these sensing devices. The BS1 will gather the content networks resources are used efficiently and will also ease their
and compute the cumulative temperature of the entire build- scalability.
ing and will raise an alarm when it observes abnormalities.
What would be the most efficient way to collect this infor-
mation that can ensure maximum utilization of the available C. Communication Protocol
resources in the constrained devices in SN1? All the use cases present a different form of communica-
With the help of the above mentioned use cases we show tion either query/response (Q/R) (1&3) or Pub/Sub (2&4).
the need for a protocol translation between Internet and sensor In many sensor networks, there is a BS that collects the
network and among sensor networks running different proto- data sensed by the sensors in the network. Another com-
cols, the lack of information availability about the nature of mon pattern is a BS controller (BSC) controls the sensor
the sensor network by the users in the Internet, the differ- devices by sending control messages to them. We can notice
ent kinds of contents requested by the users from Internet that, unlike IP, the users in these networks are interested
and other sensor network and the issue of efficient utiliza- in the content similar to ICN irrespective of their location.
tion of the resources to gather information within a sensor With an efficient communication protocol we can ensure effi-
network. Although we use these four use cases here to derive cient utilization of the constrained resources in the sensor
the requirements for ISI, we can derive many more require- networks. E.g., in use case 2&4, the user U2 and BS can
ments and even different requirements based on the nature of use Pub/Sub to collect the data periodically. Additionally,
the sensor network used as a representative. However, these with ICN in Internet we can benefit with the caching as it
basic use cases can be applied to all sensor networks and ISI not only reduces the traffic in the core network, but will
can easily support additional use cases. also reduce the traffic entering the resource constrained IoT
networks. This ensures maximum availability of the con-
tent even when the sensing devices are sleeping to save
III. R EQUIREMENT A NALYSIS resources.
In this section we build on our use cases to derive
the various requirements for integrating the sensor networks
with the Internet. We assume that sensor networks are D. Mobility
operating with a lighter version of the ICN protocol In addition to the identified usecases, another and a more
like CCN-lite (or NDN-lite) while the Internet operates important aspect to consider is the mobility of sensor devices.
with the full ICN stack like complete NDN stack. We The IoT networks should also be able to handle the mobility
choose these architectures as representative for ICN, similar of sensor devices from one domain to another. Mobility may
requirements will apply to other ICN architectures like affect the naming, reachability, and other aspects of the moving
MobilityFirst [16], etc. devices.
494 IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 2, APRIL 2018

publishers in both the Internet as well as the IoT network.


The network is also capable of delivering control messages to
the sensor networks.
2) Sensor Networks: There are many possible sensor
networks that represent numerous applications. The sensor
networks usually contain hybrid devices. The devices can
have wireless or wired access. They can be operating for
the purpose of monitoring, controlling, etc. As discussed
earlier, in this paper we propose the sensor networks to oper-
ate with the ICN protocols. We will use a lighter version
Fig. 2. ISI architecture. of the CCN called CCN-lite [8] for the sensor networks.
The sensor networks usually contain a BS or BSC that
collect the data sensed by other devices in the network
E. Security for monitoring purpose or to control their operation. The
Security is a greater concern in many if not all of the sensor BS/BSC are usually powerful machines and not constrained in
networks and Internet. There has been a growing concern that resources.
IoT’s are designed without addressing many of the associated 3) Gateway: This is the most essential piece of the
security concerns [17]. However, security induces additional proposed design. The gateway sits between the Internet run-
overhead especially in the sensor networks. There is a need ning the NDN protocol and the sensor networks running
to analyze and provide some security measures that meet the CCN-lite protocol. All of the traffic to/from the Internet and
security requirement of the sensor networks and are not an sensor networks has to pass through the gateway. The main
overkill with regards to their constrained resources. function of gateway is to perform protocol translation between
the two networks. The gateway is also responsible for mapping
IV. NDN A RCHITECTURE functions (discussed shortly). It is clear that since the IPV6
To proceed further it is important to briefly discuss the ICN MTU is 1280B while the IEEE 802.15.4 can support only
architectures we use as representative in this paper. 127B, the authors in the IETF standard [2] suggest that header
The CCN forwarding engine model contains three main compression in IPV6 is unavoidable. However, the content
components: 1) the content store (CS); 2) the pending interest generated in the sensor network is assumed to be small. We
table (PIT); and 3) the forwarding information base (FIB). believe that with efficient designs of ICN names in the sensor
The NDN communication protocol usually begins with a network the gateway does not have to compress the headers.
user interested in some content generating an interest with Although we speak about one gateway between the Internet
the respective ContentName. When the interest arrives at a and each sensor network, there is no restriction on the num-
CCN router in the forwarding engine of the router the CS ber of gateways. As the traffic exchange between the Internet
is checked to see if the content with the same name already and sensor networks increases, the burden on a single gateway
exists in which case the content is returned. In case of a CS also increases. Hence, multiple gateways have to be used to
miss PIT is checked to see if an interest with the same name distribute the load.
has already been forwarded. If yes, then the incoming face
B. Description
of the interest is recorded. If not, an entry is added to the
PIT and the FIB is checked to find the forwarding face and As said earlier we assume that the Internet is running the
the interest is forwarded. When a data packet arrives, the CS NDN protocol. There are many different sensor networks that
is checked, and if a matching entry is present then the data represent various applications like environmental monitoring,
packet is discarded otherwise the CS stores the data and PIT smart houses, etc. that are running the CCN-lite protocol. The
is checked. If an entry is found in the PIT then the data is users are spread across both the Internet as well as the sensor
forwarded to all the outgoing faces recorded in the PIT for networks. The aim of this design is to let the applications
this ContentName. Due to space constraint, we do not discuss operate in the sensor networks as they desire but extend their
the COPSS architecture and point readers to [5] for further availability and control by integrating them with the Internet.
reading. The design allows the users in Internet to access/control the
IoT devices using the Internet. A key to achieve this integration
V. A RCHITECTURE is through gateways.
The gateway is a powerful component that has many
In this section we describe our ISI architecture shown in roles to play. Every sensor network is associated with one
Fig. 2 for integrating the various hybrid sensor networks with or more gateways that is responsible for seamlessly inte-
the Internet running ICN protocols. grating the respective sensor network with the Internet.
The entire traffic between these two worlds will flow
A. Components through the gateway transparent to the users in both the
1) NDN Network: We choose NDN as a representative ICN networks.
architecture. In principal this could be any of the ICN archi- The gateway runs both the NDN protocol and CCN-lite
tectures. The network is capable of retrieving Content from protocol. The gateway maintains a mapping table which
ADHATARAO et al.: ISI WITH ICN 495

maps the lengthy, unbounded names form the Internet VI. NAMING S CHEMA
running the NDN protocol to their equivalent short In this section we discuss the naming schema for the IoT.
names in the sensor networks running the CCN-lite
protocol. A. Naming in IoT
The gateway uses a registration procedure for every device
in the sensor network. Each device upon entering the network The IoT devices usually come with the Ethernet technology
must register itself with the gateway. The gateway provides like IEEE 802.15.4 and Bluetooth LE. This results in a much
an ID to each newly added device. The device then registers smaller MTU (127B) compared to the traditional layer-2 tech-
the short name and long name of the content that it wishes to nologies adopted in the Internet. This raises a concern on the
serve. These entries will be added to the mapping table main- size of the packets that traverse the IoT network. One way
tained in the gateway. The mapping table should be updated to reduce the packet size is to have smaller names that are
whenever there are any changes in the content served by the relevant to the IoT networks. Since the any length, unbounded
sensor network. hierarchically structured names defined by many ICN archi-
We will use the term inbound traffic for the traffic enter- tectures do not suit the IoT networks. The names should be
ing the sensor network and the term outbound traffic for precise to serve the purpose of the application and yet be
the traffic going out of sensor network. The inbound traf- specific and small.
fic from the users can be either a request for data or a We describe our naming schema using the same example
request containing a control message. There are two possi- of the smart city with sensor networks as shown in Fig. 2.
ble outbound traffic. One containing the reply to the inbound Let us consider the same application of monitoring the tem-
traffic and the other is the request/subscription traffic gener- perature of all the rooms in a smart building from our use
ated inside the sensor network. The two types of outbound cases in Section II. We know each room in the smart building
traffic should be distinguished form one another as the reply is equipped with temperature sensing sensor devices and BS
traffic needs a name change through a mapping table lookup. collects the data generated by the sensing devices.
This can be achieved by using any one bit available field in the We propose a naming structure of the form
packets. Metric/ID/Area/Date/Time. The first component Metric
When the gateway receives inbound traffic it is basically an specifies what kind of data is generated by the sensing
NDN interest packet. The gateway scans the mapping table to device, e.g., temperature, pressure, humidity, etc. The second
find the equivalent short name. The gateway creates a CCN-lite component ID indicates an identifier assigned to the device.
interest packet with the short name and forwards it to the sen- The third component Area specifies the location/geographic
sor network. Upon receiving a CCN-lite data packet from the range covered by the sensing device, e.g., room, building,
sensor network, the gateway performs a lookup in the map- GPS location, etc. The fourth component Date specifies
ping table to find the long name and creates an NDN data the day at which the readings are measured and the fifth
packet with the long name, extracts the content from the CCN- component Time specifies the time at which the reading was
lite data packet and inserts it into the NDN data packet and captured by the device. The granularity of each component
forwards it in the Internet. can be application dependant, e.g., the time can vary from
Another type of out bound traffic is basically a CCN-lite hours to seconds to minutes or more. Applying this naming
interest for content located either in the Internet or in other schema to our smart building example the temperature sensed
IoT network. The gateways also support inter IoT network by a device with the id 01 in the room1 on November 3,
communication. When an IoT network running for a cer- 2016 at 12:30 could be retrieved with a name /temp/01/r1/03-
tain application needs information from outside its network, 11-16/12:30. This name is only 26B, leaving the rest of the
it simply generates a CCN-lite interest and forwards it to the packet for the content.
gateway. Since this is an out bound request traffic the gate- In most sensor networks the BS usually collects the data
way simply translates it to an NDN interest and forwards periodically from the sensing devices in the network. To dis-
it in the network. It eventually reaches the intended pub- tinguish the data retrieved from the BS we can use a naming
lisher in Internet or the gateway associated with the target schema of the form /Metric/BS/Area/Date/Time. Note that the
sensor network. When the publisher is located in Internet, component Area in the naming schema for IoT represents indi-
it follows the standard NDN protocol and reply with the vidual rooms, whereas it represent the whole building in case
data packet. If the interest reaches a gateway associated of the BS. So a user in NDN network can request for tem-
to another sensor network, the gateway performs a map- perature of the whole building or for the individual rooms.
ping table lookup to fetch the equivalent short name and This naming schema is fairly general and similar names can
prepares a CCN-lite interest and forwards it to the sensor be created for different IoT networks based on specifics of the
network. Upon receiving the data packet it prepares the NDN area or application of the IoT devices used in the network.
data packet as explained earlier and forwards it to the
Internet. The data packet finally reaches the gateway of B. Naming in Internet
the sensor network that initiated the request. The gate- Recent ICN proposals, such as NDN [4] and COPSS [5]
way performs a mere protocol translation and generates the adopt human-readable, hierarchically structured names CDs.
CCN-lite data packet as described earlier and forwards it to the Continuing our smart building example, a possible
sensor network. name structure for renaming the content for the Internet
496 IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 2, APRIL 2018

could be /temperature/UNI/ComputerScience/Building1/03- can not only subscribe to the content published inside the
11-16/12:30 for the temperature of the computer science Internet but also by the IoT networks.
building in the university on November 3, 2016 at 12:30
and /temperature/UNI/ComputerScience/Building1/room1/03-
11-16/12:30 for the temperature of room1 at 12:30. We allow C. Communication Protocol for IoT
the availability of data sensed in every room in Internet as There are two possible scopes of communication that can
some rooms might be sensitive to temperature and would have take place in IoTs: one is within the sensor networks and the
to be monitored, e.g., the temperature of a server room. other is with Internet/other sensor networks. The nature of
communication in these network differs and can greatly affect
the design choices of an efficient communication protocol for
VII. C OMMUNICATION P ROTOCOLS
the sensor networks.
In this section we discuss the current communication proto- In sensor networks, the devices are usually sensing the
cols and propose the suitable communication protocol for IoT respective data periodically and the BS gathers the sensed
and CCN network. data and analyzes it to take appropriate actions. The Pub/Sub
mode of communications seems desirable in this scenario. The
A. Query/Response Communication devices can behave like publishers and publish the data period-
This is a dominant mode of communication in current ically. While the BS can be a subscriber that subscribes to the
networks. A user interested in some content simply queries data published by all the devices in its sensor network. This
the network. Any producer of the data (or also a node with design choice allows the resource constrained sensor devices
an available copy of the data in case of ICN) responds to the to save the battery by waking up only to sense and publish
request. the data. We agree, that there are chances for the packet to get
1) Q/R Communication in IoT: In sensor network the lost. So, for reliability reason we propose the BS to resume to
BS can query the sensing devices to retrieve the sensed Q/R to retrieve only the data that was lost during the next peri-
data. The BS simply generates a request with the respective odic cycle when the sensing device is awake to sense the next
ContentName and forwards it to the network. Upon reaching reading. However, during Q/R we expect the sensor device
the producer, the sensing device responds with the requested to wait for an acknowledgment before going back to sleep
content. The BS can also send control messages using the Q/R to ensure the BS has received the packet this time. This also
mode, where the query can contain the control command while means the sensing device should retain some of its previously
the response can contain an acknowledgment of the action sensed data, which we believe can be configured based on the
taken. The IoT devices can also query for content located in applications requirement, e.g., three most recent readings for
Internet or other IoT networks. The gateway should assist in reliability.
retrieving the content in this case. In the Internet, we see that Q/R is a dominant form of
2) Q/R Communication in Internet: A user in Internet gen- communication. However, for communicating with the sen-
erates an ICN request with the ContentName of the desired sor networks we can choose between Q/R and Pub/Sub based
content and forwards it in to the network. If the content is on the need of the user. If the user (or any other application
located in the caches of any intermediate forwarding node then or sensor network) is interested in periodically receiving all of
the cached copy is returned to the user. Otherwise the interest the data collected by the sensor networks, then Pub/Sub looks
eventually reaches the producer who replies with the requested like an ideal choice in this scenario. To reduce the burden on
content. If the producer is located in an IoT network, then the resource constrained sensing devices in the sensor network,
gateway will assist in retrieving the content however, this will we believe the caching can be enabled on BS. Other sens-
be transparent to the user. ing devices can choose to retain or turn off caching based on
their available resources. Since the BS gathers the data from
all the sensors, it can act as a publisher to the subscribers in
B. Publish/Subscribe Communication the Internet. Please note that the subscribers in the Internet
In a Pub/Sub scenario, there are two roles to play: 1) pub- will subscribe to the longer names and not the shorter names
lisher and 2) subscriber. The publishers usually generate some used for publishing inside the sensor networks. So the gateway
data that could be of interest to subscribers. The subscribers has to create new publication data packet with their equivalent
maintain a long term subscriptions to the content published long names from the mapping table and then forward it to the
by the publishers (refer to COPSS [5] for detail). Whenever users in Internet.
a piece of content is published by the publishers, the network Another and most likely a common scenario is when users
will deliver it to all of its subscribers. in the Internet are interested in some particular data generated
1) Pub/Sub Communication in IoT: In IoT the sensing in the sensor networks or would like to send some control
devices that periodically sense the data can take up the role of message to the sensor networks. Intuitively, Q/R seems ideal
publisher and publish the sensed data periodically. The BS is for this scenario. When the user is interested in some data they
the subscriber interested in these contents and will subscribe just generate a CCN interest with the longer content name and
to these contents and hence will receive them when published. the network forwards it to the gateway. The gateway then per-
2) Pub/Sub Communication in Internet: Similar to IoT, forms a protocol translation and generates a CCN-lite interest
publishers and subscribers exist in ICN too. The users in ICN with the equivalent short name from the mapping table and
ADHATARAO et al.: ISI WITH ICN 497

BS1 has received the temperature of each room in the building,


it will calculate the temperature of the building and publish
a CCN-lite data with the name /temp/BS/Building1/Date/Time
(the parameter Date and Time should be replaced with the
respective values). When the GW1 receives this packet it
will check its mapping table and generate a CCN publica-
tion packet with the equivalent long name and the content and
forward it to the subscribers in the Internet.
Scenario 3: The BS1 in the SN1 needs some content say
the GPS location of the mathematics building in the smart
city. It generates a CCN-lite interest with the ContentName
/GPS/Location/UNI/Mathematics/Building2/ and forwards it
to GW1. Please note that since this is an outbound traffic
Fig. 3. Smart city use cases with ISI architecture. for content located outside the sensor network this should be
indicated by setting any available bit in the interest packet.
For CCN-lite the EXCLUDE field can be used to indicate
forwards it to the BS/BSC. Based on the type of the request this. Upon receiving this interest, the GW1 identifies it as an
received the BS/BSC either replies with the requested data or outbound traffic by inspecting the one bit field. The gateway
an acknowledgment for the action taken. will only perform a protocol translation by generating a CCN
interest with the same name and forward it to the Internet.
The content may be located in the Internet or also in another
VIII. U SE C ASE R EALIZATION sensor network. The publisher either from the Internet or any
In this section we discuss the functionality of each com- other sensor network (with the help of its gateway similar to
ponent of the proposed architecture in greater detail with the scenario1) will reply with the content. Upon receiving a data
help of the use cases we defined in Section II. packet the GW1 will again perform a protocol translation by
Consider the example network topology of a smart city with extracting the content from the CCN data packet and gener-
ISI architecture as shown in Fig. 3. There is a BS in each ate the CCN-lite data packet with the content and forward it
sensor networks that collects the temperature sensed by the to BS1.
devices periodically every 30 min. Each network is attached Scenario 4: In our scenario4 we consider the internal opera-
to a gateway that runs both the NDN and CCN-lite protocols tion of SN1. The sensor devices are sensing temperature every
for interconnecting the sensor networks and the Internet. 30 min. They are the publishers and will publish a packet
Scenario 1: A user (U1) in the Internet is interested every 30 min with the name /temp/id/room_no/Date/Time, e.g.,
in the temperature of room1 in the building1. So, U1 /temp/01/r1/03-11-16/12:30. The BS1 has subscribed to the
will generate an NDN interest with the name /temperature/ prefix /temp and hence will receive the data sensed by all the
UNI/ComputerScience/Building1/room1/03-11-16/12:30. The sensors.
network will forward the interest to Gateway1 (GW1). GW1 There are chances for the packets to get lost. If there is any
will check its mapping table for the name. When no match is packet loss, then the BS1 will generate a CCN-lite interest
found, the GW1 behaves like a router and the NDN module with the specific name. E.g., if the packet from the S2 was
running inside the GW1 will handle this packet by forward- lost then the BS1 will send a CCN-lite interest with the name
ing it to the appropriate router in the Internet. If there is a /temp/02/r2/03-11-16/12:30 at the next 30min cycle when the
match, the GW1 will generate a CCN-lite interest with the S2 will wake up to sense the next temperature reading. For
equivalent short name /temp/01/r1/03-11-16/12:30 and for- reliability reasons we require the sensors to store their latest
ward it to the CCN-lite module which will forward it to three readings. When S2 receives an interest, it will reply with
the BS1. Since the BS1 periodically collects all the data the requested content. We suggest the device S2 to stay awake
the BS1 will generate the CCN-lite data with the requested for at least 1RTT for the acknowledgment from BS1 to ensure
content and forward it to GW1 otherwise it is forwarded to that BS1 has received the data this time.
the device S1 in the SN1. Upon receiving a CCN-lite data,
the gateway will scan its mapping table to find the short
IX. D ISCUSSION
name. If there is a match the GW1 will extract the con-
tent from the CCN-lite data packet, generate a CCN data A. Mobility
packet with the equivalent long name with the extracted con- Mobility has become a norm in today’s world and IoTs
tent and forward it in the Internet. When there is no match will be no exception to fulfill this need. A simple example
found in the mapping table the GW1 will discard the data is when a smart car in a smart city moves from building1
packet. to building2. During this mobility, the smart car is detaching
Scenario 2: A user (U2) in the Internet is interested in from the sensor network in building1 and is attaching to sensor
receiving the temperature of the computer science build- network in building2.
ing every 30 min. U2 will subscribe to the CD /tempera- In order to support mobility, the gateway must handle the
ture/UNI/ComputerScience/Building1/. Every 30 min, when devices that move from one domain to another. The gateway
498 IEEE INTERNET OF THINGS JOURNAL, VOL. 5, NO. 2, APRIL 2018

can either offer a time to live (TTL) during registration and/or Li et al. [20] experimented with two ICN architectures
offer a de-registration process. During de-registration the smart MobilityFirst and NDN for IoT. They name it MF-IoT and
car sends an interest with the de-registration request to gate- NDN-IoT and compare their performance. Whereas in [21],
way and the gateway responds with an acknowledgment. The Chen et al. proposed to use ICN for IoT to realize service ori-
car can choose to wait for the acknowledgment or not. When ented communication. They use MobilityFirst as an example
the car moves to another domain it again registers itself with ICN architecture and modify it to support the service oriented
the associated gateway of the new domain. Even if the de- communication in IoT.
registration packet was lost, since the car has registered to Routing/Caching: Baccelli et al. [12] discussed the short-
anther gateway the network will synchronize with the routing comings of CCN protocol for IoT and propose a routing
updates. protocol with O(1) and almost no control traffic. They exploit
the caching and data path in ICN to support the IoT require-
B. Security ments. They also show that CCN-lite uses 80% less memory
compared to IP. Whereas in [13], Quevedo et al. stud-
1) Security in IoT: IEEE 802.15.4 provides the capability
ied the benefit of caching with ICN for IoT in terms of
for some link-layer security. The authors in IETF standard [2]
energy consumption and bandwidth utilization in comparison
urge users to make use of it. A majority of the sensor devices in
with IP.
sensor networks are expected to operate within their networks.
Protocol: Amadeo et al. [11] focused on a specific type of
Acknowledging resource constraints in the IoT devices, we
data retrieval pattern called multisource data retrieval. They
believe they should be equipped with the minimum level of
say the current NDN architecture does not support this type
security features necessary for their operation. The asym-
of communication. In the proposed solution consumers use
metric key encryption is computationally complex for the
multisource interest to retrieve data from multiple producers.
sensor networks [18], so we suggest the devices in the sensor
They propose to delete the PIT entry based on parameter like
networks can use the features provided at the link layer for
interest life time (TTL) instead of deleting when the data is
encryption and if additional security is desired then opt for
received the first time. Whereas in [22], Dinh and Kim stud-
symmetric key encryption.
ied the potential for using the ICN-based solutions for wireless
Moreover, the sensor networks will benefit with the content-
sensor area networks (WSAN). They discuss about how ICN
based security provided by the ICN solutions. Interestingly,
for WSAN’s is different from ICN for Internet. They use
Malik et al. [19] discussed attribute-based encryption (ABE)
flat names and continuous interest to receive data sensed by
for ICN networks. However, the current ABE solution are
multiple sensors as multiple sensors in WSAN sense the same
heavy for the IoT networks. The Internet on the other hand
data and respond to the interest.
can benefit greatly with ABE while the gateway can assist in
Securtiy: Compagno et al. [23] proposed a protocol for
encryption and decryption of the content using light-weight
authenticating and authorizing new devices joining IoT mesh
security measures suitable for IoT networks.
networks in ICN. They show 87% improvement in com-
2) Security in Internet: The devices in Internet are subject
munication and 66% improvement in energy consumption
to more attacks compared to devices in the sensor networks.
compared ZigBee-IP solutions. While Enguehard et al. [18]
Moreover, the devices in Internet are relatively powerful com-
compared two approaches based on asymmetric and symmetric
pared to the devices in sensor networks. These devices are
key encryptions for deploying new IoT devices in existing ICN
capable of handling complex computation and hence can
deployments. They report that although the asymmetric key-
opt for asymmetric key encryption. Although it is compu-
based solutions incur lower traffic they impose higher demands
tational heavy, it is harder to decipher the content. Many
on energy and time consumption.
ICN solutions provide security by securing the content unlike
securing the communication link as in IP. E.g., NDN uses
the digital signature of the publisher for authenticating all XI. C ONCLUSION
the content and also uses encryption for protecting private
We started with a discussion on IoT and their immanent
content.
explosive growth in near future. We discussed the short-
comings of current IoT designs and an introduction to ICN.
X. R ELATED W ORK We observed that ICN is more suitable for supporting IoT
In this section we discuss some works focused on using ICN compared to the IP architecture. We also observed that IoT
for IoT and broadly classify them into four categories: 1) archi- devices do not need the full NDN stack and can work with
tectural; 2) routing/caching; 3) protocol; and 4) security. lighter versions like CCN-lite/NDN-lite. We discussed in detail
Architectural: Amadeo et al. [14] proposed an initial high- the importance and requirements for incorporating sensor
level design for IoT using NDN architecture. They divide the networks into the Internet, thus paving a way for them to join
NDN layer into two planes: 1) data plane and 2) manage- the IoT family. We analyzed various requirements for such an
ment and control plane. The data plane handles Q/R while architecture to integrate the sensor networks and proposed ISI
the control plane re-engineers the current NDN routing plane. architecture with gateways. We described in detail the respon-
Shang et al. [3] analyzed the current TCP/IP solutions for sibilities of such a gateway. We further proposed a naming
supporting IoT. They argue that existing TCP/IP solutions are schema and communication protocol along with some pos-
inefficient and propose that IoT can benefit by using the ICN. sible mobility and security considerations for IoT networks.
ADHATARAO et al.: ISI WITH ICN 499

With the help of use cases we described the functionality of [21] J. Chen et al., “Exploiting ICN for realizing service-oriented commu-
ISI architecture. As part of future work we intend to develop nication in IoT,” IEEE Commun. Mag., vol. 54, no. 12, pp. 24–30,
Dec. 2016.
and demonstrate a working prototype of the proposed ISI [22] N.-T. Dinh and Y. Kim, “Potential of information-centric wireless
architecture. sensor and actor networking,” in Proc. Int. Conf. Comput. Manag.
Telecommun. (ComManTel), 2013, pp. 163–168.
[23] A. Compagno, M. Conti, and R. Droms, “OnboardICNg: A secure pro-
R EFERENCES tocol for on-boarding IoT devices in ICN,” in Proc. ICN, Kyoto, Japan,
2016, pp. 166–175.
[1] G. Mulligan, “The 6LoWPAN architecture,” in Proc. ACM 4th Workshop
Embedded Netw. Sensors, Cork, Ireland, 2007, pp. 78–82.
[2] G. Montenegro and N. Kushalnagar, “Transmission of IPv6 packets over
IEEE 802.15. 4 networks,” Internet Eng. Task Force, Fremont, CA, USA,
RFC 4944, Sep. 2007.
[3] W. Shang, Y. Yu, R. Droms, and L. Zhang, “Challenges in IoT Sripriya Srikant Adhatarao received the M.Sc. degree in computer science
networking via TCP/IP architecture,” Univ. California, at Los Angeles, from the Institut für Informatik of Georg-August-Universität Göttingen,
Los Angeles, CA, USA, Tech. Rep. NDN-0038, 2016. Göttingen, Germany, in 2015, where she is currently pursuing the Ph.D. degree
[4] V. Jacobson et al., “Networking named content,” in Proc. CoNEXT, in information centric networking with the Computer Networks Group.
Rome, Italy, 2009, pp. 1–12.
[5] J. Chen, M. Arumaithurai, L. Jiao, X. Fu, and K. K. Ramakrishnan,
“COPSS: An efficient content oriented pub/sub system,” in Proc. ANCS,
2011, pp. 99–110.
[6] J. Chen, M. Arumaithurai, L. Jiao, X. Fu, and K. K. Ramakrishnan,
“SAID: A control protocol for scalable and adaptive information Mayutan Arumaithurai received the master’s degree from the Technical
dissemination in ICN,” in Proc. ICN, Kyoto, Japan, 2016, pp. 11–20. University of Hamburg–Harburg, Hamburg, Germany, in 2006, and the
[7] S. S. Adhatarao, J. Chen, M. Arumaithurai, X. Fu, and Doctoral degree from the University of Göettingen, Göttingen, Germany,
K. K. Ramakrishnan, “ORICE: An architecture for object resolu- in 2010.
tion services in information-centric environment,” in Proc. LANMAN, He was with Nokia Siemens Networks, Munich, Germany, and the
Beijing, China, 2015, pp. 1–6. University of Göettingen. He was a Research Scientist with the Network
[8] CCN-Lite. Accessed: Jul. 1, 2017. [Online]. Available: Laboratories, NEC Europe Ltd., Heidelberg, Germany. He is currently a Senior
http://www.ccn-lite.net/ Researcher with the University of Göettingen. His current research interests
[9] Z. Shelby and C. Bormann, 6LoWPAN: The Wireless Embedded Internet, include IoT, future Internet (e.g., information centric networking, software-
vol. 43. New York, NY, USA: Wiley, 2011. defined networking, and network function virtualization), congestion control,
[10] Z. Sheng et al., “A survey on the IETF protocol suite for the Internet and emergency services.
of Things: Standards, challenges, and opportunities,” IEEE Wireless
Commun., vol. 20, no. 6, pp. 91–98, Dec. 2013.
[11] M. Amadeo, C. Campolo, and A. Molinaro, “Multi-source data retrieval
in IoT via named data networking,” in Proc. ACM 1st Int. Conf. Inf.
Centric Netw., Paris, France, 2014, pp. 67–76.
Dirk Kutscher received the Doctoral degree from the Universität Bremen,
[12] J. Quevedo, D. Corujo, and R. Aguiar, “A case for ICN usage in IoT
Bremen, Germany.
environments,” in Proc. IEEE Glob. Commun. Conf., Austin, TX, USA,
He is the CTO for virtual networking and IP with Huawei’s German
Dec. 2014, pp. 2770–2775, doi: 10.1109/GLOCOM.2014.7037227.
Research Center, Munich, Germany, where he is responsible for next-
[13] J. Quevedo, D. Corujo, and R. Aguiar, “A case for ICN usage in IoT
generation programmable network infrastructure research and develop-
environments,” in Proc. IEEE Glob. Commun. Conf., Austin, TX, USA,
ment. He is co-chairing the IRTF Research Group on information-centric
2014, pp. 2770–2775.
networking. He was the Chief Researcher of networking with NEC
[14] M. Amadeo, C. Campolo, A. Iera, and A. Molinaro, “Named data
Laboratories Europe, Heidelberg, Germany. He was a Visiting Researcher
networking for IoT: An architectural perspective,” in Proc. IEEE Eur.
with KDDI Research and Development Laboratories Inc., Tokyo, Japan,
Conf. Netw. Commun. (EuCNC), Bologna, Italy, 2014, pp. 1–5.
and a Researcher with the Universität Bremen. He was in leading posi-
[15] L. Zhang, D. Estrin, J. Burke, V. Jacobson, and J. Thornton, “Named
tions on several EU collaboration projects and is advising several H2020
data networking (NDN) project,” Palo Alto Res. Center, Palo Alto, CA,
projects. His current research interests include network programmability and
USA, Tech. Rep. NDN-0001, 2010.
the development and evolution of Internet protocols and architectures.
[16] D. Raychaudhuri, K. Nagaraja, and A. Venkataramani, “MobilityFirst:
Dr. Kutscher served on Technical Steering Committee of the Linux
A robust and trustworthy mobility-centric architecture for the future
Foundation OPNFV project. He is a member of the Internet Research Steering
Internet,” ACM SIGMOBILE Mobile Comput. Commun. Rev., vol. 16,
Committee
no. 3, pp. 2–13, 2012.
[17] J. Singh, T. Pasquier, J. Bacon, H. Ko, and D. Eyers, “Twenty security
considerations for cloud-supported Internet of Things,” IEEE Internet
Things J., vol. 3, no. 3, pp. 269–284, Jun. 2016.
[18] M. Enguehard, R. Droms, and D. Rossi, “Poster: On the cost of secure
association of information centric things,” in Proc. ICN, Kyoto, Japan, Xiaoming Fu received the Ph.D. degree in computer science from Tsinghua
2016, pp. 207–208. University, Beijing, China, in 2000.
[19] A. M. Malik, J. Borgh, and B. Ohlman, “Attribute-based encryption on a He was a Research Staff Member with the Technical University of
resource constrained sensor in an information-centric network,” in Proc. Berlin, Berlin, Germany. He joined the University of Göttingen, Göttingen,
ICN, Kyoto, Japan, 2016, pp. 217–218. Germany, in 2002, where he has been a Full Professor and the Head of the
[20] S. Li, Y. Zhang, D. Raychaudhuri, and R. Ravindran, “A compara- Computer Networks Group since 2007. His current research interests include
tive study of MobilityFirst and NDN based ICN-IoT architectures,” architectures, protocols, and applications for networked systems, including
in Proc. 10th Int. Conf. Heterogeneous Netw. Qual. Rel. Security information dissemination, mobile networking, cloud computing, and social
Robustness (QShine), 2014, pp. 158–163. networks.

Das könnte Ihnen auch gefallen