Content-Centric Wireless Networking: A Survey

Marica Amadeo, Claudia Campolo, Antonella Molinaro, Giuseppe RuggeriUniversity Mediterranea of Reggio Calabria - DIIES Department

Email: {marica.amadeo, claudia.campolo, antonella.molinaro,giuseppe.ruggeri}@unirc.it

Abstract

Content-Centric Networking (CCN) is a candidate future Internet architecturethat gives favourable promises in distributed wireless environments. The latterones seriously call into question the capability of TCP/IP to support stable end-to-end communications, due to lack of centralized control, node mobility, dy-namic topologies, intermittent connectivity, and harsh signal propagation con-ditions. The CCN paradigm, relying on name-based forwarding and in-networkdata caching, has great potential to solve some of the problems encountered byIP-based protocols in wireless networks.

In this paper, we examine the applicability of CCN principles to wirelessnetworks with distributed access control, different degrees of node mobility andresource constraints. We provide some guidelines for readers approaching re-search on CCN, by highlighting points of strength and weaknesses and reviewingthe current state of the art. The final discussion aims to identify the main openresearch challenges and some future trends for CCN deployment on a large scale.

Keywords: Content Centric Networking, Mobile Ad Hoc Networks, WirelessSensor Networks, Vehicular Ad Hoc Networks

1. Introduction

Modern mobile devices such as smartphones, laptops, and tablets, enabledwith wireless Internet connectivity and sensing capabilities, are steadily growingin popularity and market penetration. They can provide users with mobilityand flexibility in accessing and generating information anywhere (e.g., in home,office, shops, cars) and at any time. Wireless networking is expected to playa crucial role in the future Internet, not only to sustain direct interactionsbetween personal users devices, but also as a means to provide connectivityon a large scale while involving resource-constrained devices like sensors andsmart objects. Conventional networking protocols designed to support stableend-to-end communications between nodes that are uniquely identified throughan IP address, fail in wireless distributed environments due to dynamic changesin the network topology caused by the node mobility, frequent link failures orthe presence of energy-constrained nodes running out of battery.

Preprint submitted to Computer Networks June 3, 2014

In addition, it is also evident that the traditional host-centric Internet modelmismatches the dominant information-centric usage of the current Internet.Today, applications such as video downloading, file sharing, social networking,and cloud services, massively drive content retrieval and dissemination in theInternet. To support the efficient and reliable delivery of such applications, anumber of research initiatives has recently advocated a shift from the traditionalInternet networking model to a novel paradigm that considers the content (orinformation) as the first class network citizen and decouples it from the identityof the node(s) storing it.

Information-centric networking has become one of the main potential archi-tecture of the future Internet and several related projects are active worldwide[1]. In this research arena, the Content-Centric Networking (CCN) architectureproposed in the seminal work of Van Jacobson [2] has rapidly gained consen-sus and it is now at the basis of many research initiatives running worldwide,including Named-Data Networking (NDN) [3] and others cited in [4].

In CCN, each piece of data is associated with a location-independent namethat is directly used by the applications for content search and retrieval. Com-munication is driven by the receiver, which uses an Interest packet to request acontent by name. The content source, or any other network node that tem-porarily stores the requested content, replies with a Data packet that containsthe named content and additional authentication and data integrity informa-tion. Each Data packet is a self-identifying and self-authenticating unit; andthis enables seamless in-network caching and content replication.

It is the authors convincement that CCN is an effective networking paradigmthat well matches the features of wireless environments. Indeed, CCN can over-step the inefficiencies of TCP/IP in handling node mobility, unreliability ofwireless links, and resource-constrained devices by relaxing the need of creat-ing and maintaining stable sessions between end-points. Moreover, CCN mayleverage the broadcast channel nature and help the content sharing betweenneighbouring nodes.

Some good surveys have addressed information-centric solutions [1], [4] andcovered topics ranging from naming to mobility management and caching, e.g.,[5], [6], [7]. This paper differs from the previous ones since it focuses specif-ically on the CCN paradigm, and it provides a comprehensive overview anda clear identification of the applicability, potentialities, weaknesses and futurechallenges of this paradigm in wireless networks.

The rest of the paper is organized as follows. In Section II, we introduce theCCN basics and major functionalities. In Section III, we present the main fea-tures of wireless sensors, mobile and vehicular ad hoc networks and we identifythe benefits of CCN in such environments. In Sections IV-X different aspects ofthe CCN applicability to wireless environments are analyzed, including naming,routing and forwarding, caching, security, and transport issues, as well as eval-uation platforms and prototypes. Section XI summarizes the open challengesand future perspectives. Section XII concludes the paper.

2

Figure 1: CCN hourglass and node architecture.

2. The CCN architecture

The CCN model [2] provides a new network architecture that supports con-tent retrieval in the future Internet by using Interest/Data packets exchange.

Each CCN name is persistent, unique and hierarchical and it can be repre-sented as a Uniform Resource Identifier (URI). Integrity and authenticity aresupported at a packet-level by piggybacking the data publishers signature andother authentication information (e.g., publisher public key digest) in the Datapacket.

Since each Data is a self-contained unit, caching is facilitated in networknodes. Depending on local constraints and policies, a subset (or all) of thenetwork nodes can cache contents and speed up data retrieval while reducingthe overhead. A CCN node that maintains a cached copy of the content can actas a provider like the original source.

As shown in Figure 1, CCN inherits the hourglass model of the IP archi-tecture, but the narrow waist leverages names of content chunks instead of IPaddresses for data delivery.

Each CCN node maintains three data structures: (i) a Content Store (CS)for temporary caching of incoming Data packets; (ii) a routing table namedForwarding Information Base (FIB) used to guide the Interests towards Data;and (iii) a Pending Interest Table (PIT), which keeps track of the forwardedInterest(s) that are not yet satisfied with a returned Data packet.

Routing in CCN serves the purpose of computing the FIBs entries to be usedfor Interest forwarding. Given the hierarchical name structure, CCN facilitatesglobal routing via prefix aggregation.

The CCN forwarding plane is a two-step process that involves Interests for-warding from the consumers to the retrieved data, and Data packets flowingback along the same path to the consumers. Each CCN node receiving an In-terest makes its forwarding decision based on the following algorithm. First, itsearches for a name prefix longest-match in its CS. If a match is found, thenthe node sends the Data back to the incoming interface of the processed Inter-est. Otherwise, if there is a matching PIT entry (another consumer has alreadyasked for the same Data), the Interest is discarded and the new incoming inter-face is added to the existing PIT entry. Otherwise, a new PIT entry is createdand the Interest is further forwarded to the interface stored in the FIB.

3

Figure 2: Interest/Data packets processing and forwarding operations in CCN.

When a Data packet is retrieved, its name is used to look up the PIT. Ifa matching entry is found, then the node sends the packet to the interface(s)where the Interest was received, it stores the data in the CS, and deletes the PITentry. So, Data packets follow the chain of PIT entries back to the requester(s).If a match is not found in the PIT then the Data packet is considered unsolicitedand it is dropped.

For the sake of clarity, Figure 2 sketches CCN packets processing and for-warding. Upon receiving the Interest from node A, the intermediate node C,not finding a match in its CS or in the PIT, forwards the Interest to the sourcenode D. Once receiving the Data packet from D, node C forwards it back toA, and subsequently it serves directly the request for the same content comingfrom B with its cached copy.

CCN achieves one-to-one flow balance by letting each Interest be consumedby a single Data packet. Moreover, it permits to specify different transportservices at the Strategy Layer, depending on the application requirements (suchas reliability, delay-tolerance) and the network constraints (such as mobility,channel quality).

3. Content-Centric Wireless Ad Hoc Networking

3.1. Main Features of Wireless Ad Hoc Networks

Wireless ad hoc networking can be regarded as a type of spontaneous infras-tructureless networking, automatically activated when nodes are in line of sightwithout the need of any centralized control. It can be characterized by differentdegrees of node mobility, multihop communications, battery-powered devices,and multifaceted possible deployments and use cases.

Mobile Ad hoc NETworks (MANETs) are self-organized multihop networksthat support exchange of information without relying on any pre-existing net-work infrastructure. Applications cover home/office environments, tactical net-works, emergency services. A MANET can be used either as a stand-alone

4

deployment with locally generated and exchanged data, or to provide wirelessInternet access through gateway(s) connected to the infrastructure. A MANETcan be formed among users devices sharing similar interests (e.g., students ex-changing class materials in a campus; workforce operators exchanging maps ina disaster recovery scenario), or sharing location-based information (e.g., com-muters exchanging information about train/bus departures times).

Unlike MANETs where routing nodes are mobile, in wireless mesh networks(WMNs) routers are stationary and form a wireless multihop backbone [8].Mesh routers offer wireless connectivity to mobile devices that may use themesh backbone to connect to the Internet through one or more gateways. Amesh network can benefit from advance planning of the node positions, butnothing prevents it from growing organically.

The ad-hoc networking paradigm is also at the basis of Vehicular Ad hocNETworks (VANETs). By enabling vehicle-to-vehicle and vehicle-to-infrastructurecommunications, VANETs can provide a unique set of applications specificallydesigned for the road environment to improve safety and comfort of drivers andpassengers, e.g., by disseminating hazardous event notifications, road traffic in-formation, advertisements about nearby points-of-interest [9]. Mobile nodes ina VANET move at higher speed than nodes in a MANET and with predictablemovements. The high node mobility may cause low connectivity and highlydynamic network topology with frequent partitions. Vehicular nodes are en-abled with self-localization capability, and have resources not limited by energy,memory, or processing constraints.

The lack of an infrastructure is among the main features of Wireless SensorNetworks (WSNs), which consist of (thousands of) resource-constrained devicesthat communicate untethered [10]. These networks are used for tasks such asenvironmental monitoring, logistics, surveillance; hence they cannot be operatedin isolation but need to be connected to remote servers. Sensor devices are themost critical in terms of energy, memory and processing resources.

Table 1 summarizes the main features of the aforementioned networks. Al-though they have a huge potential in different scenarios and arouse interestfrom service/network providers and users, they suffer from serious technicalchallenges that could hinder their massive deployment and efficient use.

Wireless channel. Signal propagation on the wireless medium may beadversely affected by impairments like interference, path loss, multipath fading,and shadowing effects, which could induce packet errors and losses.

Distributed control. The broadcast wireless channel may facilitate datasharing on the one hand, but on the other hand it asks for specific channelaccess policies to keep collisions and packet redundancy under control. Thedistributed channel access in most wireless networks is based on carrier senseand may suffer from hidden and exposed terminals problems with throughputdegradation, especially harmful in multihop dynamic scenarios.

Mobility. Network topologies dynamically change due to node mobility,which can range from low-to-medium (e.g., in a MANET) to high mobility(e.g., vehicular nodes). Topology changes may cause network partitions andlead to poor, intermittent, and short-lived connectivity with negative effects on

5

Table 1: Main features of Wireless Ad hoc Networks.

Feature MANETs VANETs WMNs WSNs

Mobility Medium High Low-to-Static Medium-to-Static

Battery con-

straints

Medium (de-vices arerechargeable)

No constraint(energy istaken from theengine)

Low-to-Noconstraints(nodesare mostlyplugged)

High

Storage capa-

bilities

Medium-to-Low

Theoreticallyinfinite

High Low

Main reference

standard

IEEE802.11a/b/g/n

IEEE 802.11p IEEE 802.11s IEEE 802.15.4

the routing performance.Constrained Resources. With the exception of infrastructured elements

and nodes on board of vehicles, wireless nodes are battery-powered deviceswith limited processing power and storage capabilities. These constraints areespecially critical for sensors.

3.2. CCN advantages in wireless environments

Node mobility, multihop communications, battery constraints, the lossy broad-cast wireless channel, the type of applications, and the lack of an infrastructurecharacterizing wireless ad hoc networks, heavily question the capabilities oftraditional TCP/IP protocols to support efficient and robust end-to-end com-munications [11]. This is why, over the years, alternative networking solutionshave been devised that try to move away from host-centric models and embracecontent-oriented communications.

Early precursors of the content-centric paradigm can be found in the lit-erature on wireless networking (e.g., delay-tolerant [12] and opportunistic [13]networking, content-based publish-subscribe models [14]). These solutions areimplemented as an overlay on the IP layer, which is used to address the networknodes. Unlike them, the named-data CCN architecture can be also implementedon top of any layer 2 access technology as a pure clean-slate solution.

The feasibility of applying CCN in wireless ad hoc scenarios, such as general-purpose [11] and military [15, 16] MANETs, VANETs [17, 18], and WSNs [19],[20], [21] has been recently discussed in the literature, with preliminary deploy-ments in some cases [22], [23], [24]. The motivations for such a surge of interestare manifold.

First, consumer mobility is intrinsically supported in CCN: when a consumermoves, it can simply re-issue any unsatisfied Interest from the new location.Provider mobility may require routing updates, but CCN inherently supportscontent multi-sourcing, thus reducing the effects of a provider re-location [6].

Second, CCN retrieves information without the need of any a priori knowl-edge of the source node identity. This is a clear benefit for several mobile appli-cations that are information-centric in nature: uploading a photo to Facebook

6

or Twitter, downloading videos from YouTube are examples of the commonway for mobile users to access the Internet. Other emerging applications arerelated to the exchange of surveillance data, command and control, and soft-ware updates among mobile users in a MANET. Similarly, in VANETs, traffic,weather, and parking information can be requested by vehicles in a given area,regardless of their identities or IP addresses. A variety of civilian, scientific andmilitary applications based on data collection and dissemination in large-scalemonitoring sensor networks, can also benefit from hierarchical content namingand the simple CCN Interest/Data exchange.

In addition, the majority of these applications consists of information ad-dressed to more than one recipient. Such information can be created explicitlyfor public dissemination (e.g., news, weather information), or it can involve re-stricted groups of recipients (e.g., a video streaming), or it can be increasinglygenerated by end users (e.g., a post in a social network). CCN natively sup-ports multicast data delivery, thanks to the Interests aggregation in the PIT,according to which intermediate nodes avoid forwarding multiple requests forthe same Data packet while the first one is pending.

The fourth major advantage is that CCN can cope well with intermittent,short-lived connectivity, and dynamic topologies in wireless ad hoc environments.In fact, under node mobility, low power operation and opportunistic contacts,having self-consistent data units and exploiting in-network decentralized datacaching/replication can substantially improve the quality of communication bymaking the best of the broadcast wireless medium.

In the following sections, the key design challenges of the CCN paradigm arediscussed in detail, by surveying solutions proposed in the literature. For thereaders convenience, the most representative application domains for wirelessnetworks, their main demands and native CCN benefits are summarized in Table2. In addition, Table 3 provides a summary of the main works that, to the bestof our knowledge, recommend the adoption of CCN in MANETs, VANETs andWSNs. Most of them propose general-purpose solutions without any specificapplication in mind.

The majority of the surveyed literature supports CCN as a clean-slate so-lution. An overlay of CCN on IP is not generally recommended in ad hoc net-works for two main reasons: (i) the end-to-end route set-up and maintenancebetween overlay nodes induce high control overhead; (ii) the overlay designforces point-to-point communications, without exploiting neither the broadcastradio channel nor in-network caching [25].

4. Naming

A CCN content name is composed of one or more variable length alphanu-meric strings separated by / , e.g., a Youtube video name can be /youtube/-clipNetworking/CCN/introduction. CCN defines some basic conventions for thehierarchical name structure (e.g., encoding human-readable name components,globally-unique name prefixes), while the name semantics and the number ofsubstrings in a name can be customized on the basis of applications, local and/or

7

global conventions. As a consequence, application developers can choose a hi-erarchy of name components that fits their needs and lets name conventions tobe opaque to the network [47].

The CCN naming system is still under active research, and some namingproposals in the context of wireless networks have recently started to appear.

In [17] the authors explore the benefits of hierarchical CCN naming inVANETs. The following name structure is proposed for traffic information dis-semination: /traffic/geolocation/timestamp/datatype, in which the name com-ponents identify the temporal and geographical scopes of traffic information,and the application data type. For instance, the Interest with name /traf-fic/Road101/south/40,41/ could be used to request traffic information abouta specified region of Road 101 (southbound, kilometres 40-41). Similarly, in[24], the interested road area is encoded in the name; e.g., /traffic/westwood-at-strathmore/ would refer to the traffic information from the area close to theWestwood-Strathmore street intersection.

In [21], CCN names are customized to support sensor networking. In orderto fit into an IEEE 802.15.4 frame, the authors assume that the maximumlength of a content name is 50 octets and limited to five components withmaximum 15 octets each one. In [19] a naming scheme for WSNs is proposedto describe the sensing task, thus allowing the sink to precisely ask for theneeded information, and the sensors to describe the sensed data. The namestructure task type/task location/task time period/nonce accounts for: thesensing task (e.g., temperature, humidity); the geographic area in which thetask is performed, stated either in terms of logical names (e.g., a room) orin geographical coordinates; the time period in which the task is performed

Table 2: Application domains and their requirements coupled with CCN benefits and relatedliterature.

Application domain Main requirements CCN benefits Works

Battlefield operations anddisaster relief (MANETs)

High security, self-configuration, resiliency

Data integrity and originauthentication, possibil-ity of encryption, mean-ingful naming, multipathsupport

[15], [16]

Vehicular safety applica-tions (VANETs)

Timely and reliable high-priority safety messagesdelivery

Broadcasting, meaning-ful naming, robust trans-port

[26]

Road traffic efficiency andinfotainment applications(VANETs)

Scalable delivery oflocal/spatial-relevantinformation

Lightweight route set-up and maintenance,caching

[17], [18],[24], [27],[28], [29],[30], [31],[32], [33]

Environment/Buildingmonitoring (WSNs)

(Large scale) period-ical short-lived smalldata delivery, energyefficiency

Thin naming, easy con-figuration, Interest ag-gregation

[19], [20],[21], [22]

Video streaming(MANETs, VANETs,WMNs)

High bandwidth, low la-tency

Caching [34], [35]

8

Table 3: Literature Addressing CCN on top of Wireless Ad Hoc and Sensor Networks.

MANETs VANETs WSNs

Naming - [17], [24] [19], [21]

Routing and for-

warding

[11], [15], [25], [36],[37], [38], [39], [40],[41], [42]

[18], [27], [28], [29],[30], [33]

[19], [21]

Transport [25], [37] [27], [31] -

Caching [25], [43] [27] [20], [21]

Security [15], [16], [44] [32] -

Prototypes [39], [45], [46] [24] [21], [22], [23]

Service models [26], [27] [20]

(e.g., an instantaneous measurement, or an averaged value over a given timewindow); the unique data identifier used to identify replicas. For instance,an Interest with name humidity/room121/[timestamp1,timestamp2]/1323454declares that the sink is looking for the average humidity in room121 during thetime period between timestamp2 and timestamp1.

In summary, the CCN namespace is highly expressive and highly customiz-able. By leveraging the hierarchical tree structure, CCN name components canbe user-friendly attributes that describe the content itself.

5. Routing and forwarding

Unlike in IP, where routing is the smart operation and forwarding is con-sidered as dumb, in CCN both routing and forwarding are smart. Routingrefers to the way FIBs are populated by exchanging name-prefix announcementsamong routers; forwarding refers to Interest and Data processing, which is donehop-by-hop according to the decisions of the Strategy Layer in each node.

CCN nodes are expected to use any of the traditional routing protocols(adapted to handle content names instead of IP addresses, e.g., like in [48]) tofill in the FIB tables and keep them up to date. So far, the definition of proac-tive routing protocols in CCN wireless ad hoc networks has not been specificallyinvestigated. This is mainly because to manage name-prefix advertisements isvery challenging in mobile networks and may introduce some overhead (e.g.,related to data source mobility, dynamic data catalogues, possibility of aggre-gating name prefixes, frequency of updates, etc.), so that the cost of maintainingrouting information may overwhelm the benefits of proactive solutions [40]. Inaddition, in such distributed environments contents may be time-and location-relevant and be generated on the fly. Therefore, the CCN literature usuallyrelies on reactive flooding-based approaches to discover content in wireless net-works. Only a recent work [33] proposes proactive advertisements in case ofpopular non-sharable/non-cacheable data, by using Bloom filters to reduce theoverhead.

The CCN stateful forwarding plane may leverage information stored in thePIT and the FIB to make forwarding decisions adaptive to network conditions

9

[49]. Nonetheless, the design specifics of the CCN forwarding fabric that fitwireless networks and applications remain to be filled. Strategy modules couldbe customized that sometimes violate the basic CCN forwarding rules.

We organize the remainder of this section in two parts. First, we scan simpleflooding-based solutions for data dissemination in content-centric wireless net-works. Then, we examine enhanced techniques that introduce some selectivityin the CCN forwarding decision process by leveraging additional (discovered)information about the neighbourhood and the producer(s). We call blind andaware the two forwarding approaches respectively.

5.1. Blind forwarding

CCN implementations in wireless environments may leverage the broadcastnature of the radio channel to help data dissemination [11], [27], [37], [42].

Flooding is the easiest way to forward Interest packets on the wirelessmedium. Such an approach has the virtue of simplicity and well faces situa-tions in which end-to-end path set up and maintenance are difficult and costly,such as in dynamic ad hoc environments and with resource-constrained devices.Flooding facilitates content sharing in the network; in fact, a node overhearingsome data of interest requested by other nodes can access it without an explicitrequest. This reduces the number of transmissions and saves the nodes en-ergy. However, flooding on a broadcast medium must be handled with care andcontrolled to avoid the broadcast storm [50].

To counteract packets redundancy and collisions, solutions in the literaturemainly rely on distributed packet suppression techniques. The basic idea is thata node defers the packet forwarding while overhearing the channel and, eventu-ally, drops the packet if it hears the packet transmitted by a neighbour [11, 15].Distance-based, slotted random, or purely random defer strategies can be imple-mented. In [27] a set of timers is used to assist Data broadcasting in vehicularenvironments. Specifically, a collision-avoidance timer is used by neighbour-ing cars that simultaneously receive an Interest for traffic jam information, toreschedule Data broadcasting at different times. A similar approach is followedin [18, 30], where different defer timers are used for Interest and Data forwardingin order to minimize the collision probability and prioritize Data over Interests.

However, a blind controlled flooding based on the above mentioned simplecountermeasures does not always guarantee that (i) the best nodes are selectedto forward packets, and that (ii) overhearing avoids packet collisions. This iswhy controlled flooding can be regarded as a baseline implementation of CCNbroadcasting in wireless networks, on top of which more sophisticated and awareforwarding strategies can be deployed, as discussed in the following.

5.2. Aware forwarding

New awareness mechanisms have been included in the forwarding plane tohelp in selecting the outgoing interface, the content provider(s), and the next-hop nodes, by leveraging new entries in (new) tables, additional packets and/oradditional fields piggybacked in Interest/Data packets.

10

Interface selection. CCN forwarding may leverage the information storedin PITs and FIBs to select the outgoing interface(s) at each node. For example,the FIB may keep track of the delivery performance (e.g., in terms of latency,throughput, round-trip times, cost) of each outgoing interface, so that packetsare transmitted via the best performing interface [49]. In [26], vehicles thathave access to multiple networks (such as IEEE 802.11p, WiMAX, UMTS)transmit safety messages over the low latency interface. Packets could be alsosimultaneously transmitted over all available interfaces of different technologiesto cope with disruption in connectivity [24].

As a further option, the outgoing radio interface can be selected so as toooad the cellular infrastructure by leveraging the ad hoc connectivity of nearbynodes. This is the approach followed in [34] and [35] for mobile video streaming,where the CCN routing-by-name is used to enforce the download of a videosegment either through cellular or Wi-Fi interfaces.

Next-hop selection. Some awareness can also be used by a CCN node toselect the next-hop in Interest/Data forwarding. In the direction-selective datadissemination solution in [41] the Interest sender initially divides its surroundingspace into four quadrants and broadcasts the Interest (including its own node-idand geo-location information) to all one-hop neighbours. The farthest node ineach quadrant is then selected as a relay node for the Interest packet. In [25]the eligibility of a relay node is decided by its data retrieval rate for the givenname prefix and its distance to the data consumer. If the data retrieval rate islow or the node is too far away from the consumer, the incoming Interest willbe discarded locally. In [27] a pushing timer is used in vehicular networks toforward locally generated data (e.g., an accident warning) further away fromthe point it was originated (e.g., towards drivers travelling towards the accidentplace). A farther car from the previous transmitter uses a shorter timer than anearby car to schedule Data re-broadcasting.

The BlooGO proposal [38] determines if forwarding the packet or not bycomparing the neighbourhood of the sender and the receiver. This is possiblesince a Data packet carries a Bloom filter field that includes the identifiers ofthe nodes in the transmission range of the sender. They are collected throughperiodical beacon broadcasting. A node forwards the packet only if its localneighbourhood is not completely included into the one advertised in the incom-ing packet, so that the progress of packets is ensured without much redundancy.BlooGO is used as the routing protocol in the MADN platform [39], a modulararchitecture for multipath data distribution in content-centric MANETs.

In [19] a forwarding strategy is designed that creates a direction state duringthe initial content discovery phase in a WSN. After receiving an Interest, thediscovered producer sends the Data and includes its identifier in an additionalpackets field. A receiver node stores this identifier in the so-called Next HopTable (NHT) and then forwards the packet towards the consumer with its ownidentifier information. As a result, the NHT in each crossed node contains abind between the content name, the producer, and the next hop identifier and apath is created that is used any time the consumer sends a subsequent Interestto retrieve more Data.

11

Path selection. Thanks to the broadcast nature of the wireless medium,Interests may propagate along multiple paths towards potential provider(s), andeven Data packets may be returned over multiple paths, especially when mul-tiple copies of a content item are cached in the network. Multipath retrievalis particularly beneficial in wireless ad hoc networks, because it can mitigatethe service disruption periods due to node mobility or adverse propagation con-ditions, and can limit the overhead caused by end-to-end path establishmentand maintenance. In [29], the data diversity over multiple paths is beneficiallyexploited in a VANET. By applying network coding techniques, intermediatenodes perform a linear combination of the received chunks, and then they trans-mit the result to the neighbours.

Provider selection. If the consumer discovers more than one contentsource, the best performing provider can be selected based on some criteria.

In [15], after an Interest reception, a provider transmits a Reply packet toadvertise itself in a MANET. The consumer may collect more than one Replyfrom different providers, select one of them and send back a Request to the targetprovider, which is allowed to reply with Data. Similarly, in [41] four packets(Interest, ACK, CMD and Content) are exchanged. A three-way handshakescheme is also proposed in [11] with the similar intent of routing Data over themost stable consumer-provider path.

The selected provider can be advertised in subsequent Interests so that in-termediate nodes can properly route the packet, as advocated in [30, 28]. Lightpath-state information (the selected providers identifier and its hop distance tothe consumer) are included in Interest and Data and left as bread crumbs in aProvider Table kept by CCN nodes in a MANET. In doing so, the twofold bene-fit is achieved of keeping the channel load under control and reducing downloadtime and energy consumption [37]. One could reasonably argue that by fixingonly one provider could reduce the CCN advantages of fetching the content fromdifferent nodes. However, in-network caching is fully operative, in the sense thatin case of packet losses, any intermediate node caching the missing Data cancompensate for the loss and provide it, although not being the selected provider.

6. Caching

On-path data caching provided by CCN is especially useful in infrastruc-tureless wireless networks: it promises high content availability, network trafficreduction, and low retrieval latency, by mitigating the challenges induced bylossy links, bandwidth limitations, and node mobility. By caching data, a mo-bile node may also enable store-carry-and-forward communications and serve asa link between disconnected areas. This particularly suits the VANET environ-ment [27].

Caching in ad hoc networks has been widely studied in the literature thatpreceded CCN, especially in the context of opportunistic networking and datamuling. However, the novelty of CCN is the coupling of caching and named-data. In fact, names make the content accessible in an application-independentmanner, so that a request for a named content can be satisfied by any matching

12

data regardless of its location. Additional peculiarities of the CCN architecture(e.g., splitting content in chunks, correlated content requests) make caching inCCN a quite new, and widely uninvestigated, research topic, as briefly discussedin the following.

Chunks partitioning. Unlike most existing works, where entire objects aregenerally cached, in CCN content is partitioned in chunks of small size, so thatdifferent chunks of the same object may be cached on different CCN nodes. Insuch a case, the cached fragment phenomenon discussed in [25] may affect dataretrieval, especially if caching is coupled with a single-path forwarding strategythat keeps sending Interests to the first found content source, like in [36]. Infact, although the complete data object resides in a single provider (i.e., a singlephysical node), consumers may wrongly select other discovered nodes that holdonly partial objects. To overcome this issue, the node sending Data shouldadvertise if it owns the entire object or only some chunks.

Another issue to consider is that CCN requests for consecutive chunks ofthe same object are correlated, so that the traditional independent referencecaching model no longer holds [51]. Although few recent studies consideredcorrelated arrivals, the analysis is limited to simple topologies with single paths(e.g., cascade or tree) and, hence, they cannot be straightforwardly extended toad-hoc dynamic topologies.

Cache decision and replacement policies. Cache decision shall be takenat each CCN node regarding whether or not to cache the current data; then, incase of positive decision and if the cache is full, the node may need to replace astored chunk according to a replacement policy. The CCN architecture does notmake specific assumptions on cache decision and replacement policies; howeverthe related literature has typically considered that all nodes may cache all newchunks (this is the Leave a Copy Everywhere, LCE, assumption), and that theLeast Recently Used (LRU) chunk is replaced if a cache is full.

Concerning caching decision, caching every content in every node along thedelivery path(s) may cause caching redundancy, so playing against the CCNefficiency if not coupled with a smart forwarding. Furthermore, indiscrimi-nate caching may waste network bandwidth and device energy, due to mul-tiple transmissions of cached contents from many nodes; and this is a threatin wireless environments. In order to select what to cache (and then eventu-ally what to replace), in the traditional caching literature, popularity-drivenapproaches are considered to favour keeping more popular contents in caches.In [43], the Location-Aided Content Management Architecture (LACMA) forMANETs binds data to geographic locations and more densely replicates popu-lar contents so to push them closer to potential consumers. However, as arguedin [51], also forcing many replicas of popular content in multiple caches may bedetrimental for cache diversity in the CCN context with correlated arrivals andnave forwarding on multiple paths.

Beyond the decision about what to cache, the dynamicity of many wire-less networks, characterized by short-lived contacts, and the spatial- and time-relevant nature of locally exchanged contents (e.g., in a VANET) make thedecision about where to cache to be a major concern in mobile networks. While

13

caching in vehicular devices comes at a negligible cost, since on board units arenot limited by energy and storage constraints, the same consideration does nothold for resource-constrained devices like sensors [21, 20]. The authors of [27]propose a data muling service in VANETs where each vehicle caches overhearddata, even though it is not interested in, and then performs proactive data push.

Solutions where only selected nodes in the delivery path cache the contentmay leverage the node betweenness centrality in a topology, according to whichsome nodes have higher probability of getting a cache hit [52]. Centrality-baseddecisions could be easily applied in static wireless scenarios (such as a meshbackhaul), but they are more difficult to extend to mobile scenarios, where thetopology dynamically changes and the concept of node centrality is less intuitive.

7. Security

Despite the clear benefits of content-centric location-independent security,many issues still need to be tackled. Some of them, including the possibility ofcache pollution and Denial of Service (DoS) attacks via Interest flooding, arecommon to wired and wireless networks [16]; while others are strictly related towireless environments, e.g., the computational cost of data-security in presenceof resource-constrained devices, the absence of trusted third parties in infras-tructureless scenarios. So far, CCN security in ad hoc and sensor networks hasbeen poorly investigated; most of the work is focused on (or at least testedin) wired topologies, e.g., [53]. The main reason is that the wired environmentis easier to tackle until some robust solutions have emerged and matured. Inthe following, we focus on the few fundamental security aspects that should beconsidered in wireless-specific design.

The problem of key distribution and management in tactical and emergencyMANETs has been tackled in [15] and [16]. The authors assume that the con-ventional CCN security framework is enabled, but public and private keys mustbe pre-assigned before nodes are dispersed in the field by using a pre-definedkey management tool. It is widely accepted, in fact, that several key manage-ment schemes (with fully or partially distributed certificate authority) designedin past years for ad hoc networks [54] can be extended to CCN nodes.

A new alternative approach to verify the public-key and producer identitybinding in wireless networks is presented in [44], where a social network-basedsecurity scheme is proposed that employs a trusted chain of friend relation-ships. Since the producer identity is included in the Data packet, the contentrequester will first look up into the local Identity Bundle Table (consisting ofthe identity-id and its public key). A match implies instant verification. If thelocal table does not contain this binding relationship, the requester must sendout another Interest packet with the identitys name to retrieve the producersidentity bundle from the social trust graph. By doing so, both authenticity andintegrity problems are solved in a more flexible and distributed way.

In [32], a secure application is built for data collection from vehicles thatallows manufacturers to verify integrity and authenticity of incoming contentand to protect privacy of mobile users. Data packets originated by vehicles are

14

tagged with their signature and encrypted using the public key of the referencedatabase server. The authors assume that the data collector has access to eachmobiles public key. This is reasonable for vehicle manufacturers as they couldrecord public keys of vehicles and also store the public key of the database serverinside vehicles before release.

The burden of security support in memory-constrained devices has not beenconsidered in preliminary implementations of CCN in WSNs [20, 21, 22, 23],where the Data packet is assumed to include only Payload and Content namefields. Similarly, the cost of security operations in terms of time and energyconsumption has not been analyzed in presence of battery-powered ad hoc andsensor devices. However, it would be worth exploring in depth security issuesin resource-poor wireless nodes to better figure out their actual impact on CCNperformance, which could heavily change when authentication and cypheringprocedures will be in place.

8. Transport

The CCN Strategy Layer may perform some functions that are typical ofthe Internet transport layer, e.g., unacknowledged packet retransmissions andrate regulation. Differently from the TCP, these functions are implemented byCCN nodes hop-by-hop and not end-to-end.

Interest retransmissions. Due to the shared and lossy nature of thewireless channel, Interests/Data packets may be lost or corrupted in transit, orData may be temporarily unavailable due to the provider mobility. To support areliable transport, if a pending Interest is not satisfied in a given period of timewith a returned Data packet, a new Interest must be retransmitted. The relatedretransmission timeout (RTO) setting is critical to quickly recover packet losseswhile limiting useless retransmissions.

Currently, a specific algorithm for the computation of the Interests RTOin CCN networks has not been defined. In [27], where traffic information dis-semination is VANETs is considered, every vehicle broadcasts a packet severaltimes with a pre-configured RTO. When the node hears that the packet hasbeen successfully re-broadcasted, it cancels subsequent retransmissions.

TCPs RTO estimation [55] has been extended to work also in content-centricVANETs [31]. Each CCN node tracks the time when an Interest has been for-warded and records a Round Trip Time (RTT) sample when the requested Datais received. Then, the average RTT is estimated as a moving average of RTTsamples, and the RTO is dynamically adapted to follow the average RTT varia-tions. The problem is that the dynamics of ad hoc networks topologies coupledwith the channel unreliability and potential congestion may create fluctuationsin such estimation. In addition, different nodes can store the same content intheir caches; so a consumer could receive successive Data packets from differentnodes. As a consequence, the RTT fluctuations could be very high. To copewith this issue, it is crucial for the consumer to know the identity of the contentsource, in order to maintain separated RTT measurements and/or to perform

15

selective updates. In [37], for example, the RTT estimation is updated only ifthe Data packet is sent from the selected provider.

Interest rate regulation. In CCN multiple Interests asking for successiveData may be pipelined to maximize the bandwidth usage. By properly tuningthe Interests transmission rate, a node can control the traffic flow accordingto the available network resources. Interest rate control is however still poorlyinvestigated in the literature for wireless ad hoc networks.

In [25], the Neighborhood-Aware Interest Forwarding (NAIF) uses localstatistics to adjust the fraction of Interests a node in a MANET should for-ward for a given name prefix. NAIF is based on the following intuition: themore Data of the same name prefix a node overhears from its neighbors, themore Interests corresponding to that name prefix it can drop.

In this way, the nodes cooperatively regulate the Interest forwarding ratewithout congestion.

Another Interest rate control scheme is presented in [37], where a transportfunction is defined for wireless multihop environments. The proposed mecha-nism adapts the Interest transmission at the consumer-side on the basis of theobserved Data arrival rate and an explicit feedback from intermediate nodesthat advertise the minimum sustainable data rate on a given path. The re-ceived Data rate at a consumer gives an indirect measure of global congestionin the network; while the sustainable data rate gives information on the localcongestion in a node (the bottleneck) on the path. The Interest rate is regulatedso to be slightly higher than the received Data rate, while not overloading anynode in the path.

9. Overhauling the CCN philosophy

In the previous sections, literature solutions have been surveyed that pro-pose enhancements to the main pillars of the CCN model, caching policies,while keeping the main tenets of the paradigm, i.e., receiver-driven communi-cation supporting both one (source)-to-one (consumer) and asymmetrical one(source)-to-many (consumers) communication modes. However, for the sake ofcompleteness, it is worth observing that some works also proposed some revi-sions of the CCN philosophy to enable not natively supported service models.

With its Interest/Data exchange, CCN natively supports a pull (or on-demand) service model, where the consumer starts communication by declaringthe requested content and there is a 1-to-1 relationship between Interest andData. Nevertheless, with proper adaptations, push (or publish-subscribe) ser-vices, in which Data packets are sent without any Interest solicitation, could alsobe supported by CCN [57]. This could be the case of media streams or real-timenotifications, such as sensors immediately reporting abnormal detected param-eter values [21], [20], and cars transmitting safety messages [26] or gatheringdata about surrounding environments (e.g., traffic jam, road closure) [27].

In addition to the addressed case where information is generated by a singleprovider and requested by multiple recipients, some applications in the wireless

16

domain may involve more than one content producer. For instance, a sink nodemay be interested to gather all temperature information from many sensors ina place. In this case, a consumer sending the Interest may expect to receivemultiple content objects (with names that share some common parts) frommultiple sources.

Depending on the application domain, three major add-ons to CCN can beidentified to support the mentioned service models. Such modifications requirethe CCN communication fabric to be properly re-engineered both in terms oftransport mechanisms and semantics of packet types.

1. Pushing via Unsolicited Data. In [26], unsolicited content packets calledEvent packets are used to disseminate safety information in a VANET. TheEvent Packet has the same structure as the CCN Data, but features an addi-tional field called Expiry Time that indicates the time after which the packetshould be deleted. Similarly, in [27], unsolicited Data are published by a car atthe head of a vehicle sequence in the travelling direction and then disseminatedby other cars acting as data mules.

2. Pushing via Long-term Interests. The concept of long-term Interests hasbeen examined in [56], [57] and used in [58] to deliver multiple real-time contentpackets with only one Interest. In this implementation, Interests are not deletedafter a matching Data is forwarded, but they remain in the PIT until users ex-plicitly unsubscribe from a channel or their lifetime expires. Therefore, Interestpackets are extended with one more optional selector fields, which indicate thepacket type: long-term or normal. Such a concept could be also successfullyapplied to support periodical (untriggered) data monitoring in WSNs.

3. Multiple Data via Continuous Interests. The notion of continuous Inter-est is used in [20] to handle many (sources)-to-one (consumer) communicationmode. Similarly to long-term Interest, the continuous Interests lifetime is setfor a long period of time and the packet must not be deleted after the soliciteddata has been received, thus a node could receive the same kind of data fromdifferent producers.

10. Evaluation tools

Several simulation and emulation tools are currently available to analyze theCCN performance and its potential extensions to operate in wireless and mobilead hoc networks.

The majority of discussed works has used customized CCN modules as eval-uation tools built on top of existing simulation platforms like ns-2 [37], Qualnet[25], [36]. In order to incentivize studies on CCN a new software module forthe open-source ns-3 network simulator [59], namely ndnSIM [60], has been re-leased in 2012. An official simulation environment that strictly follows the CCNcommunication model, ensures more accurate results and the reproducibilityand the comparability of simulations conducted by the CCN research commu-nity. Under active development worldwide, ndnSIM supports the core featuresof CCN in a modular way and can be the best environment to simulate large

17

scale wireless networks. In fact, ns-3 provides modules that reproduce mobilityand propagation models, and various access layer technologies such as IEEE802.11g and 802.11s.

Due to its recent deployment, to the best of our knowledge, only a few papersusing ndnSIM for CCN performance evaluation in wireless networks have beenpublished, e.g., [19], [27].

CCNx [61] is an open source software reference implementation of the CCNarchitecture and protocol, developed at Palo Alto Research Center. It is avail-able for deployment on several operating systems such as Linux, Unix, MacOSand Android. The core component of CCNx is the ccnd daemon, which sup-ports the forwarding plane and the caching service; it currently can run as anoverlay on top of IP to take advantage of existing connectivity.

Recently, NDNBlue has been released [62], a cross platform proxy layer forLinux and Android systems, which works between CCNx and Bluetooth stacksto achieve CCN connectivity directly over Bluetooth links.

In [22], an extension of CCNx is presented to support a content-centric com-munication layer over Contiki, an open-source operating system for embeddeddevices and WSNs that relies on IEEE 802.15.4 at the physical and MAC layers.

Another fully customizable and open source platform is the CCN-Java Open-source Kit EmulatoR (CCN-Joker) for wireless ad hoc networks [63]. It is anapplication-layer platform, specifically tailored for wireless devices with limitedresources in terms of storage capability and computational load, which can beused to build a CCN overlay and is suitable for both emulation-based analysisand real experiments.

11. Open Challenges

From the literature overview in the previous Sections, it clearly emerges that,despite the young age of the topic, several works have appeared addressing CCNin wireless environments, due to its inherent potentialities. The main findingsfrom the scanned literature are summarized in Table 4 that shows the mainpotential benefits of CCN for wireless networking and the research trends foreach of the main CCN pillars.

Despite the enhancements and modifications proposed in the CCN communi-cation fabric to overstep challenges and constraints of wireless ad hoc networks,research in this field is still at the beginning and some hints for future deploy-ment can be provided as follows.

As regards naming schemes, there is a tight relationship between naming,applications and access network constraints. Although CCN names can havevariable lengths without any a priori fixed upper bound, wireless technologiessuch as IEEE 802.15.4 have very limited payload sizes and should work withthin content names. It is mandatory for application designers to interact withthe CCN developers in order to converge on some standard application-specificand access network-compliant naming definitions.

18

Table 4: CCN for wireless networking: main benefits and research prospectives.

CCN pillar Main benefits Research prospective

Naming (i) Low-cost network configu-ration (ii) Theoretically infi-nite/unbounded namespace

Naming schemes adapted to applicationstype, institution requirements, networkconstraints, and/or global conventions

Security (i) Content-based security; (ii)No need of securing chan-nels/boxes in the delivery path

(i) Key management infrastructures; (ii)Computation- and bandwidth-efficientsignature schemes; (iii) Effective and flex-ible trust models

Routing and

Forwarding

(i) Lightweight route setup andmaintenance; (ii) Easy mul-ticasting; (iii) Multipath for-warding and multiple providers;(iv) Leveraging broadcastingand channel overhearing

(i) Advanced controlled flooding and re-active schemes; (ii) Prioritization policiesfor different types of contents; (iii) Net-work coding techniques to enhance mul-tipath routing performance, (iv) Robustpacket suppression techniques

Caching (i) Coping with intermittentconnectivity and error pronechannels, (ii) Shortening thecontent retrieval time

Policies adapted to device constraints,content type, node and network features

Transport Connectionless communications (i) Interest/Data retransmissions proce-dures, (ii) Interest rate control policies

Many issues related to security are completely open. It is worth noticingthat many wireless nodes are resource-constrained devices and signature and au-thentication operations can be computationally expensive in terms of time andenergy resources consumption. At the same time, several applications in wire-less domains, e.g., for control systems [64], may require the use of authenticatedInterests in addition to signed Data. This further complicates the managementof the wireless security framework. Therefore, the use of public key cryptog-raphy claims for two open tasks: (i) definition of an efficient key managementmethodology that works in infrastructureless environment, and (ii) developmentof computation- and bandwidth- efficient signature schemes.

The number of proposals handling CCN routing and forwarding chal-lenges in wireless environments witness the interest of the research communityin these topics. Overall, there is a wide consensus on leveraging some kind ofawareness in the Strategy layer to augment the CCN forwarding fabric. How-ever, the trade-off between the overhead of transferring and/or keeping aware-ness in every node (e.g., additional information about providers and/or neigh-bors) and the achieved benefits in terms of packet delivery performance shouldbe carefully considered by accounting for the requirements of the applications,the nodes capabilities and the network conditions. Moreover, the routing designshould be tighten to both caching and transport routines.

Concerning caching, storage is becoming cheaper and of smaller footprint:modern smartphones and tablets have significant storage capacity often reachingseveral gigabytes. Thus, caching space would not be a big matter, unless toconsider battery-constrained devices and sensors typically equipped with a fewkilobytes memory. Several design options shall be considered to decide where,what, and how long caching data. For instance, in an environment with nodes

19

equipped with heterogeneous capabilities, data storage could be distributed in afew nodes, more powerful than the others. In a more general case, the popularity,the priority, and the type of contents could make the difference to decide whatand how long caching data.

Transport issues pose several concerns related to the regulation of the In-terest rate and the estimation of the retransmission interval, which are especiallycritical in wireless environments with dynamic topologies and high node mobil-ity. Both aspects are crucial to ensure reliability, flow balance, and congestioncontrol in distributed wireless environments.

An additional aspect to consider for a comprehensive analysis of CCN inwireless environments includes the deployment mode. Although the intro-duction of any new technology always claims for additional costs and compati-bility issues, it is worth noticing that wireless access nodes (access points, meshrouters, road-side infrastructure units, etc.) and devices (smartphones, tablets,vehicular on board units, etc.) can be easily augmented with a software CCNstack that works directly over the access layer technology, thus building purestand-alone content-centric environments. Connectivity through an IP-basedbackbone can be performed by enabling some nodes with proxy functions or byimplementing overlay solutions.

Finally, it is interesting to briefly speculate on the relationships betweencontent-centric wireless networking and other emerging paradigms, like cloudcomputing [65] and social networking [66].

On the one hand, it is worth investigating how CCN can help to (i) makecloud computing deployable on a smaller scale in mobile environments, and (ii)support mobile social networking applications. On the other hand, it should beexplored if and to which extent (i) CCN can benefit from mobile cloud comput-ing, e.g., to augment distributed in-network content storage, and (ii) the designof cross-layer protocols inspired from social networking analysis can improveCCN performance (e.g., through socially-driven forwarding and caching opera-tions) and reduce security-related threats to make content delivery trustworthy(e.g., by exploiting social relationships among nodes).

12. Conclusions

In this paper we provided a survey on the state-of-the-art of the content-centric networking principles and architecture applied to wireless ad hoc envi-ronments (e.g., MANETs, VANETs, and WSNs).

By leveraging named data, in-network caching and lightweight forwarding,the Content Centric Networking paradigm is a particularly attractive solutionfor wireless networking, facing the limitations of resource-constrained devicesand overstepping mobility and wireless channel issues, hardly addressed by con-ventional TCP/IP-based solutions. Irrespective of the huge CCN potentialitiesin wireless environments, research on this topic is still in its infancy; many re-search challenges still lie ahead, mainly concerning security and privacy, cachingand transport issues, and have to be addressed to bring CCN for wireless net-working to life.

20

References

[1] B. Ahlgren, C. Dannewitz, C. Imbrenda, D. Kutscher, and B. Ohlman, A Survey ofInformation-Centric Networking, IEEE Communication Magazine, vol. 50, no. 7, 2012.

[2] V. Jacobson et al., Networking Named Content, in ACM CoNEXT, Rome, Italy, 2009.[3] Named Data Networking (NDN) Project, http://www.named-data.net/.[4] G. Xylomenos et al., A Survey of Information-Centric Networking Research, IEEE

Communication Surveys and Tutorials, vol. 16, no. 2, 2014.[5] M. Bari, S. Chowdhury, R. Ahmed, R. Boutaba, and B. Mathieu, A Survey of Nam-

ing and Routing in Information-Centric Networks, IEEE Communications Magazine,vol. 50, no. 12, pp. 4453, 2012.

[6] G. Tyson, N. Sastry, I. Rimac, R. Cuevas, and A. Mauthe, A Survey of Mobility inInformation-Centric Networks: Challenges and Research Directions, in ACM NoM12,Hilton Head, SC, USA, June 2012, pp. 16.

[7] G. Zhang, Y. Li, and T. Lin, Caching in Information Centric Networking: A survey,Computer Networks, 2013.

[8] M. Conti and S. Giordano, Multihop Ad Hoc Networking: The Reality, IEEE Com-munications Magazine, vol. 45, no. 4, pp. 8895, 2007.

[9] G. Karagiannis, O. Altintas, E. Ekici, G. Heijenk, B. Jarupan, K. Lin, and T. Weil,Vehicular networking: A survey and tutorial on requirements, architectures, challenges,standards and solutions, IEEE Comm. Surveys & Tutorials, vol. 13, no. 4, 2011.

[10] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, Wireless sensor net-works: a survey, Computer networks, vol. 38, no. 4, pp. 393422, 2002.

[11] M. Meisel, V. Pappas, and L. Zhang, Ad Hoc Networking via Named Data, in Mo-biArch10, Chicago, Illinois, USA, September 2010.

[12] K. Fall, A Delay-Tolerant Network Architecture for Challenged Internets, in ACMSIGCOMM03, New York, NY, USA, 2003, pp. 2734.

[13] J. Scott, P. Hui, J. Crowcroft, and C. Diot, Haggle: a Networking Architecture DesignedAround Mobile Users, in IFIP WONS, January 2006, pp. 7886.

[14] F. Guidec and Y. Maheo, Opportunistic Content-Based Dissemination in DisconnectedMobile Ad Hoc Networks, in IEEE UBICOM07, November 2007, pp. 4954.

[15] S. Y. Oh, D. Lau, and M. Gerla, Content Centric Networking in Tactical and EmergencyMANETs, in Wireless days, IFIP 2010, Venice, Italy, 2010, pp. 15.

[16] B. Etefia and L. Zhang, Named Data Networking for Military Communication Systems,in IEEE Aerospace Conference, Big Sky, MT, USA, March 2012, pp. 17.

[17] J. Wang, R. Wakikawa, R. Kuntz, R. Vuyyuru, and L. Zhang, Data Naming in Vehicle-to-Vehicle Communications, in IEEE INFOCOM12 Workshop on Emerging DesignChoices in Name-Oriented Networking, March 2012.

[18] M. Amadeo and C. Campolo and A. Molinaro, Content-Centric Networking: is that aSolution for Upcoming Vehicular Networks? in ACM VANET12, June 2012.

[19] M. Amadeo, C. Campolo, A. Molinaro, and N. Mitton, Named Data Networking: aNatural Design for Data Collection in Wireless Sensor Networks, in IFIP Wireless Days,Valencia, Spain, 2013.

[20] N.-T. Dinh and Y. Kim, Potential of information-centric wireless sensor and actor net-working, in International Conference on Computing, Management and Telecommuni-cations (ComManTel), 2013.

[21] Z. Ren, M. A. Hail, and H. Hellbruck, CCN-WSN - a lightweight, flexible Content-Centric Networking Protocol for Wireless Sensor Networks, in SSNIP 2013, Melbourne,Australia, April 2013.

[22] B. Saadallah, A. Lahmadi, and O. Festor, CCNx for Contiki: Implementation Details,INRIA, Tech. Rep. RT-0432, November 2012.

[23] J. Meijers, M. Amadeo, C. Campolo, A. Molinaro, S. Paratore, G. Ruggeri, and M. Booy-sen, A Two-Tier Content-Centric Architecture for Wireless Sensor Networks, in IEEEICNP13, Gottingen, Germany, 2013.

[24] G. Grassi, D. Pesavento, L. Wang, G. Pau, R. Vuyyuruc, R. Wakikawac, and L. Zhang,Vehicular inter-networking via named data, ACM SIGMOBILE Mobile Computingand Communications Review, vol. 17, no. 3, 2013.

21

[25] Y.-T. Yu, R. B. Dilmaghani, S. Calo, M. Y. Sanadidi, and M. Gerla, Interest Propaga-tion in Named Data MANETs, in IEEE ICNC13, San Diego, CA, January 2013.

[26] G. Arnould, D. Khadraoui, and Z. Habbas, A Self-Organizing Content Centric NetworkModel for Hybrid Vehicular Ad Hoc Networks, in ACM DIVANet11, Miami, Florida,October 2011.

[27] L. Wang, A. Afanasyev, R. Kunts, R. Vuyyuru, R. Wakikawa, and L. Zhang, RapidTraffic Information Dissemination Using Named Data, in ACM NoM12), June 2012.

[28] M. Amadeo, C. Campolo, and A. Molinaro, CRoWN: Content-Centric Networking inVehicular Ad Hoc Networks, IEEE Communications Letters, vol. 16, no. 9, 2012.

[29] P. T. Fard and V. Leung, A Content Centric Approach to Dissemination of Informa-tion in Vehicular Networks, in Second ACM International Symposium on Design andAnalysis of Intelligent Vehicular Networks and Applications (DIVANet12), 2012.

[30] M. Amadeo, C. Campolo, and A. Molinaro, Enhancing Content-Centric Networking forVehicular Environments, Elsevier Computer Networks, vol. 57, no. 16, 2013.

[31] M. Amadeo and C. Campolo and A. Molinaro, Design and Analysis of a Transport-LevelSolution for Content-Centric VANETs, in IEEE ICC Workshops, Budapest, Hungary,June 2013.

[32] J. Wang, R. Wakikawa, and L. Zhang, DMND: Collecting Data from Mobiles UsingNamed Data, in IEEE Vehicular Networking Conference, December 2010.

[33] Y.-T. Yu, X. Li, M. Gerla, and M. Sanadidi, Scalable vanet content routing using hier-archical bloom filters, in Wireless Communications and Mobile Computing Conference(IWCMC), 2013 9th International. IEEE, 2013, pp. 16291634.

[34] B. Han, X. Wang, N. Choi, T. T. Kwon, and Y. Choi, AMVS-NDN: Adaptive MobileVideo Streaming and Sharing in Wireless Named Data Networking, in IEEE INFOCOMNOMEN Workshop, Turin, Italy, 2013.

[35] A. Detti, M. Pomposini, N. Blefari-Melazzi, S. Salsano, and A. Bragagnini, Ooadingcellular networks with Information-Centric Networking: the case of video streaming, inIEEE WoWMoM, 2012, pp. 13.

[36] M. Meisel, V. Pappas, and L. Zhang, Listen First, Broadcast Later: Topology-agnosticForwarding Under High Dynamics, in Annual Conference of International TechnologyAlliance in Network and Information Science, September 2010.

[37] M. Amadeo, A. Molinaro, and G. Ruggeri, E-CHANET: Routing, Forwarding andTransport in Information-Centric Multihop Wireless Networks, Elsevier ComputerCommunications, vol. 36, no. 7, 2013.

[38] F. Angius, G. Pau, and M. Gerla, BLOOGO: BLOOm filter based GOssip algorithmfor wireless NDN, in ACM NoM12, June 2012.

[39] F. Angius, A. Bhiday, M. Gerla, and G. Pau, MADN Multipath Ad-hoc Data Networkprototype and experiments, in IEEE IWCMC13, Cagliary, Italy, 2013.

[40] M. Varvello, I. Rimac, U. Lee, L. Greenwald, and V. Hilt, On the Design of Content-Centric MANETs, in IEEE/IFIP WONS, January 2011.

[41] Y. Lu, B. Zhou, L.-C. Tung, M. Gerla, A. Ramesh, and L. Nagaraja, Energy-efficientcontent retrieval in mobile cloud, in Proceedings of the second ACM SIGCOMM work-shop on Mobile cloud computing, ser. MCC 13. ACM, 2013.

[42] C. Anastasiades, A. Uruqi, and T. Braun, Content Discovery in Opportunistic Content-Centric Networks, in IEEE Local Computer Networks Workshops, October 2012.

[43] S.-B. Lee, S. H. Y. Wong, K.-W. Lee, and S. Lu, Content Management in a Mobile AdHoc Network: Beyond Opportunistic Strategy, in IEEE INFOCOM 11, April 2011,pp. 266270.

[44] Y. Lu, Z. Wang, Y. Yu, R. Fan, and M. Gerla, Social network based security scheme inmobile information-centric network, in Med-Hoc-Net, June 2013.

[45] P. Sharma, D. Souza, E. Fiore, J. Gottschalk, and D. Marquis, A Case for MANET-Aware Content Centric Networking of Smartphones, in IEEE WoWMoM, June 2012.

[46] R. Chiocchetti, D. Rossi, and G. Rossini, SCALE: a Content-Centric MANET, in IEEEINFOCOM 13, Turin, April 2013.

[47] L. Zhang et al., Named Data Networking (NDN) Project, PARC, Tech. Rep. NDN-0001, October 2010.

22

[48] L. Wang et al., OSPFN: An OSPF Based Routing Protocol for Named Data Network-ing, NDN Project, Tech. Rep. NDN-0003, July 2012.

[49] C. Yi et al., A case for stateful forwarding plane, Elsevier Computer Communications,vol. 36, no. 7, April 2013.

[50] O. Tonguz, N. Wisitpongphan, J. Parikh, F. Bai, P. Mudalige, and V. Sadekar, On theBroadcast Storm Problem in Ad Hoc Wireless Networks, in Broadband Communica-tions, Networks, and Systems (BROADNETS), San Jose, CA, October 2006.

[51] G. Rossini and D. Rossi, Evaluating CCN multi-path interest forwarding strategies,Computer Communications, vol. 36, no. 7, 2013.

[52] W. K. Chai, D. He, I. Psaras, and G. Pavlou, Cache less for more in Information-Centric Networks (extended version), Computer Communications, vol. 36, no. 7, 2013.

[53] A. Afanasyev, P. Mahadevan, I. Moiseenko, E. Uzun, and L. Zhang, Interest FloodingAttack and Countermeasures in Named Data Networking, in IFIP Networking, 2013.

[54] J. V. D. Merwe, D. Dawoud, and S. McDonald, A survey on peer-to-peer key manage-ment for mobile ad hoc networks, ACM CSUR, vol. 39, no. 1, p. 1, 2007.

[55] V. Jacobson, Congestion Avoidance and Control, in SIGCOMM88, New York, NY,USA, 1988.

[56] C. Tsilopoulos, G. Xylomenos, Supporting Diverse Traffic Types in Information CentricNetworks, PACM SIGCOMM workshop on Information-centric networking (ICN11),Toronto, Canada, August 2011.

[57] A. Carzaniga, M. Papalini, and A. L. Wolf, Content-Based Publish/Subscribe Net-working and Information-Centric Networking, in ACM SIGCOMM Workshop onInformation-Centric Networking (ICN11), Toronto, Canada, August 2011.

[58] C. Yao, L. Fan, Z. Yan, and Y. Xiang, Long term interest for real-time applications inthe named data network, in ACM AsiaFI12, Kioto, Japan, August 2012.

[59] The Network Simulator-3 (ns-3), http://www.nsnam.org/.[60] A. Afanasyev, I. Moiseenko, and L. Zhang, ndnSIM: NDN simulator for NS-3, NDN

Project, Tech. Rep. NDN-0005, July 2012.[61] CCNx Project, http://www.ccnx.org.[62] A. Attam and I. Moiseenkoy, NDNBlue: NDN over Bluetooth, Tech. Rep. NDN-0015,

2013.[63] I. Cianci, L. A. Grieco, and G. Boggia, CCN - Java Opensource Kit EmulatoR for

Wireless Ad Hoc Networks, in ACM CFI12, Seoul, Korea, September 2012.[64] L. Breslau, P. Cao, L. Fan, G. Phillips, and S. Shenker, Securing Instrumented Envi-

ronments over Content-Centric Networking: the Case of Lighting Control and NDN, inIEEE Infocom NOMEN Workshop, Turin, Italy, 2013.

[65] M. Gerla, Vehicular Cloud Computing, in IEEE Med-Hoc-Net, 2012.[66] B. Mathieu, P. Truong, W. You, and J.-F. Peltier, Information-Centric Networking:

a Natural Design for Social Network Applications, IEEE Communications Magazine,vol. 50, no. 7, pp. 4451, 2012.

23