Professional Documents
Culture Documents
Skoutas Makris Templatesurvey
Skoutas Makris Templatesurvey
Abstract—During the last years, a continuous convergence has been invented and roughly exploited in many fields of
and fusion of computing and networking systems is stressed in informatics since 1960s and refers to the idea that computers
the ICT research community. Context awareness appears to be can sense, react and possibly adapt their functionalities based
a major research ”glue-point” of such kind of heterogeneous
environments’ integration. In this survey paper, we aim to on the information they acquire from their environment. The
assemble all the ”puzzle-pieces” of the emerging research area of term ”context awareness” (CA) was first explicitly introduced
Context Aware Mobile and Wireless Networking (CAMoWiN) in in the research area of pervasive computing in [1] and refers, in
order to provide the reader a holistic representation of the work general, to the ability of computing systems to acquire and rea-
having been done in the past emphasizing in recent research son about the context information and adapt the corresponding
efforts. A precise definition of abstract Context Aware (CA)
functionalities together with a clear classification of related state- applications accordingly. During the last decade, increasing
of-the-art proposals are provided in the form of a proposed interest has been observed on ways to share and exchange
taxonomy scheme per CA functionality. Each taxonomy scheme context information among remote and heterogeneous CA
can also serve as a sufficiently general framework permitting the systems, too. Towards this trend, mobile and wireless systems
easy classification of novel CAMoWiN solutions emerging in the appear as the most promising and challenging networking
near future. Therefore, our study not only defines all the possible
components of a CAMoWiN architecture and clearly points out research area for the introduction of novel CA functionalities.
the similarities and differences of all research efforts in the field As a matter of fact, numerous mobile computing devices
but also evaluates future research trends and paves the way for and plethora of wireless networking technologies have been
the researchers to work beyond the concepts investigated in this developed so as the research area of CA computing [2]
work. Finally, apart from identifying the open challenges, which to be broadened. Conclusively, Context Aware Mobile and
arise when applying traditional concepts of pervasive computing
to mobile and wireless networks, we further map them to the Wireless Networking (CAMoWiN) is the research area, which
Mobile Cloud Computing paradigm, whose concepts appear to is surveyed throughout this work and can be defined as a
be very promising regarding next generation networking and fusion of CA computing with mobile and wireless networking
computing environments’ integration. (cf. figure 2).
Index Terms—Context Awareness, Pervasive Computing, Mo- The main driving force and over-arching principle guiding
bile and Wireless Networks, Uncertainty Management, Mobile
Cloud Computing.
our research motivation is the ways that uncertainty manage-
ment is related with CAMoWiN as well as with cross-layer
and cross-system optimization mechanisms. The notion of
I. I NTRODUCTION uncertainty in CA computing appears mainly as a consequence
of the complexity of context acquisition mechanisms and
S INCE the very early stages of computer science real-
ization, computing and networking research fields have
been jointly paving the way towards the Information and
context processing. That is, the presence of uncertainty may
harm the users’ confidence in the application rendering it use-
Communication Technologies (ICT) revolution. Even nowa- less [3]. Moreover, from the mobile and wireless networking
days, a continuous convergence and fusion of computing and perspective, uncertainty poses stiff challenges to the seamless
networking systems is stressed between both academic and functionality of ubiquitous services and applications and ap-
industrial ICT research community members. In this survey pears in various facets in all layers of the traditional protocol
paper, we focus on a related research ”glue-point” of such kind stack [4]. Therefore, cross-layer approaches, which assume
of convergence called context awareness (CA). As a result, we that any kind of information (including context) has to be
investigate CA concepts from both computing (i.e. pervasive, exchanged between different layers of the protocol stack, have
ubiquitous, etc) and networking (i.e. mobile, wireless, etc) to be adopted so as end-to-end performance can be optimized
points of view trying to balance the content provided for these by adapting each layer against this information [5]. Optimized
two high-level perspectives. Context, as a research notion, performance regarding cross-layer context information flow
among numerous CAMoWiN entities can also be achieved
Manuscript received 11 November 2011; revised 9 March 2012. as stated in [6] by classifying the various CA functionalities
The authors are with the Dept. of Information and Communication Systems in corresponding abstraction layers. Finally, given the fact
Engineering, University of the Aegean, Samos, Greece, 83200 (e-mail:
{pmakris, d.skoutas, cskianis}@aegean.gr). that next generation wireless networks’ reality conglomerates
Digital Object Identifier 10.1109/SURV.2012.040912.00180 many heterogeneous networking systems, cross-system ratio-
1553-877X/12/$31.00
c 2012 IEEE
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, ACCEPTED FOR PUBLICATION
nale helps in defining ways to exchange uncertain contextual by widely accepted standards. The section ends up with: a)
information among heterogeneous entities residing in different an accumulative table of some selective CAMoWiN imple-
systems, following the same rationale points as in cross- mentations accompanied by all possible CA functionalities,
layering regarding overall optimal system performance. providing thus the reader with a distinctive visualization of
The major purposes of this survey paper can be summed up the progress having been made during the last years and b)
to a four-fold contribution points’ package being sententiously a figure demonstrating the paper’s roadmap by assembling all
described below. First of all, we deal with a newly founded CAMoWiN’s functional puzzle-pieces. In sections IV-IX, we
research area called Context Aware Mobile and Wireless follow the same structure. Therefore, for each CA abstract
Networking (CAMoWiN), which is an up-to-date version of functionality discussed per section: a) an overview of past
the well-known CA computing area. More specifically, we contributions is given, b) today’s state-of-the-art approaches
claim that adopting traditional concepts of pervasive comput- are outlined and c) implications of future challenges are
ing and supplementing them with others dealing with mobile discussed. More specifically, sections IV-VII deal with context
and wireless networking notions, new research challenges can information acquisition, modeling, exchange and evaluation.
emerge broadening thus CA applicability. Our second main Section VIII provides means of exploiting context awareness
contribution is the proposed taxonomy scheme. By outlining from business logic perspective towards the invention of novel
the need for a precise and differentiated definition of abstract CA applications. Section IX deals with security, privacy and
CA functionalities, we provide a clear classification of related trust related issues, which are applicable in every stage of the
state-of-the-art proposals. The proposed taxonomy scheme can context information flow life cycle. Finally, the paper ends up
also serve as a sufficiently general framework permitting the with section X, which summarizes the main open CAMoWiN
easy classification of novel CAMoWiN solutions emerging challenges per CA functionality and provides implications on
in the near future. As a result, we claim that our work can ways that these challenges can be applicable in an up-to-date
significantly help researchers to understand the similarities CAMoWiN-related research paradigm called Mobile Cloud
and differences of all research efforts in the field having been Computing.
made so far and these, which are going to be made in the
near and (hopefully) mid-term future. Moreover, we believe II. T HE U SE OF C ONTEXT AND C AMOWIN E VOLUTION
that our work can be useful for both: a) naive readers and Context is a notion whose better understanding and use
researchers whose speciality resides out of the scope of this can enable the rapid evolution of CAMoWiN by providing
paper and b) professionals who seek ways and ideas in order insights into abstract mechanisms and functionalities required
to work beyond the concepts investigated in this work. The to support this field of research. Various past out-of-date and
third main contribution point refers to features outlined in a incomplete definitions have been proposed in the literature.
traditional survey paper, whose main objective is to provide Some early definitions in 1990s defined context by example
the reader a holistic representation of the work having been [1], [7] and some others tried to interrelate it with other
done in the past emphasizing in recent research efforts towards notions such as the environment or situation [8], [9] limiting
envisioning the corresponding open challenges. Finally, we thus context’s applicability range. Dey in [10] achieved to
go one step further by appropriately mapping the majority provide a definition, which proved to be adequately diachronic
of CAMoWiN’s open challenges with an up-to-date research for the initial stages of CA computing. According to [10],
paradigm’s (called Mobile Cloud Computing) strengths and ”context is any information that can be used to characterize
opportunities, justifying thus our title’s statement regarding the situation of an entity. An entity is a person, place, or
the trend towards networking and computing environments’ object that is considered relevant to the interaction between
integration. a user and an application, including the user and applica-
Conclusively, the paper’s short-term objectives can be sum- tion themselves”. This definition, apart from being extremely
marized to the fact that a complete and comprehensive frame- general, has some substantial drawbacks. Table I summarizes
work for CAMoWiN area is provided for all the research and them providing thus five rationale points that differentiate
industrial community and thus all related previous, ongoing past and up-to-date context definitions in the literature. First
and upcoming works can be mapped and well positioned of all, nowadays, context is a collection of measured and
according to the proposed taxonomy schemes. Regarding long- inferred knowledge [11] rather than just a set of values with
term perspectives, this paper provides hints for realizing the no underlying understanding of what these values ultimately
ultimate integration of diversified networking and computing mean. Secondly, there is a lack of clear separation between
ecosystems via the sententious analysis of emerging Mobile the concepts of context and context information [12]. The
Cloud Computing paradigm. latter concept implies a process of exploiting the context
The remainder of the paper is structured as follows. In sec- information in various abstract ways and a perpetual flow
tion II, CAMoWiN area is defined by: a) presenting how the of context information among local or remote heterogeneous
semantic notion and use of context have changed during the entities. Thirdly, context does not simply ”characterize the
last two decades, b) investigating ways that general uncertainty situation of an entity”. Contrariwise, context can also arise
management principles can be applicable and c) discussing from the general activity of a next generation CA system, thus
its interactions with related research areas. In section III, generating and sustaining the context [13]. Another drawback
a plethora of past and state-of-the-art CA frameworks and of Dey’s definition is that it takes for granted the fact that
architectures are surveyed, starting from early task specific ar- context exists only when ”an interaction between a user and
chitectures of 1990s to latest complex architectures supported an application” occurs. This is not true especially in the field
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 3
TABLE I
PAST VS . U P - TO - DATE C ONTEXT D EFINITION RATIONALE POINTS
Decentralization
Cross‐system & Cross layer optimization
Autonomicity
Context Heterogeneous conectivity management Mobile &
Aware Business aware network management Wireless
Computing Networking
Personalization Services
Security, privacy, trust management
Cloud Computing
CAMoWiN area
solution. On the other hand, an inconsistency may be incurred Finally, the meaning of unknown context is quite easy
if plenty of such sub-optimal decisions are made in a short to understand. In traditional CA systems, unknowns usually
time interval leading thus to violations of high-level and result from sensor failures and various connectivity problems
fixed strategies, whose existence ensure the long-term system’s [16]. On the other hand, in CAMoWiN systems, an extended
stability. concept is adopted considering unknown context as infor-
mation, which is (in whatever way) unavailable at specific
Existence of wrong context information is not an unusual
time instances. Therefore, in some cases, even though specific
phenomenon in a CAMoWiN system because much of the
context may exist somewhere in the system, it has to be
context being exchanged among heterogeneous entities is
available the exact time it is needed for use, otherwise it
delivered via unreliable wireless links. Referring to early CA
may be considered as completely unknown by the system.
systems, erroneous context information arises as a result of
So far, by investigating all possible aspects that uncertainty
human error and the use of brittle heuristics to derive high-
can appear in a general CA system, we gave the first hints
level implications from low-level data [16]. In CAMoWiN
regarding the main requirements, which are necessary for
systems, erroneous context may also be acquired by remote
managing the complicated and inherently uncertain structures
entities, thus not permitting MTs to process the received con-
of context. Moreover, we showed that the semantic notion
text. Wrong context can also appear in terms of inadequacy of
and use of context has rapidly changed during the last decade
information. One may observe that inadequacy is a synonym
following the evolution steps of mobile and wireless net-
of incompleteness. Nevertheless, inadequate context means
working. In the rest of this section, we further show how
that the information lacks of some crucial elements, which
context awareness (CA) meets some emerging next-generation
cannot be recovered or inferred by the system ultimately
networking and computing trends creating thus the need of
classifying it as wrong. Furthermore, let us assume the same
CAMoWiN research area to evolve (see figure 2).
use case described for incomplete context information earlier
in this section (i.e. when signalling messages are exchanged One of the main challenges of next-generation networking
between a wireless access network and a MT). The sense of is the lack of centralized goals and control [17]. Therefore,
inadequacy lies in the fact that there is no available time for the all recent research efforts propose either distributed system
MT to request for the same information because this would architectures or hybrid ones, where centralized components act
become out-of-date and generally useless (e.g. strictly real- as a subordinate system supplement [18]. Nowadays, various
time procedures). Not applicable context is the information, past shortcomings encompassed in hardware restrictions of
which is received by entities that are not able or should not ex- MTs such as power supply, bandwidth, processing capabilities
ploit it for reasons of overall system efficiency. In CAMoWiN and storage are assumed to be manageable with the use
systems, large amounts of heterogeneous context is delivered of next-generation cognitive devices and thus distributed ap-
in the wireless medium and thus mobile devices usually proaches can be implemented more easily [19]. Another major
receive and process useless or practically wrong messages. advantage of decentralized approaches is that they increase the
This problem becomes even greater in next generation mobile robustness of personalized services and thus the invention of
and wireless networking where cooperative communications novel or killer-applications and services at a macroscale is
field emerges as a new research trend (read more in section boosted. Therefore, today’s CAMoWiN systems consist of a
VI). variety of distributed components, which have to be both com-
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 5
plicated and well implemented in order to be able to reduce the In: Raw context data
complexity of CA applications, improve maintainability and Out: Final low-level context
TABLE II
R EPRESENTATIVE PAST & STATE - OF - THE - ART CAM OW I N IMPLEMENTATIONS
Limbo [33] P - L L - - - -
√
Context Toolkit [36] - L L - L - L
Cooltown [42] P - L - L L - -
CA middleware
approaches
CoBrA [43] P P P L L - - L
√
SOCAM [44] P L P L - - -
√
CASS [47] P P P P L - -
√
MobiPADS [48] L P P P P - -
√ √
PACE [20] P - P L - L
√
CAPP [51] P - P L P - -
CA autonomic
networking approaches
√ √ √
CA3RM-Com [52] - P P L -
√ √ √ √
E3 [24] L P L -
√ √
Tramcar [55] P P L P L -
Latest CA architectures
√ √ √
CAMMS [58] P L P L P
√ √ √ √ √
HURRICANE [22] L L P
√ √ √ √
P1900.4-based [61] L P P P
√ √ √ √ √
CASUP [28] P L -
√ √ √ √ √
FOCALE [23] P L L
√ √ √ √
C-CAST [6] L P P L
√ √ √ √ √
SemEUsE [63] P P L
architecture [40] are some more example architectures dealing assumed centralized architectures and frameworks in order
with the same issue. to overcome storage, processing and energy constraints of
traditional mobile devices. However, decentralized approaches
CA Middleware Approaches: Encapsulating the context gradually emerged along with the evolution of cognitive
management logic into middleware rationale was another MTs and the need for exchanging and disseminating context
solution for the problem of decoupling context capturing and information between remote entities. The primary goal of
context processing from application composition [41]. Tradi- middleware approaches in distributed wireless environments
tional middleware approaches emerged as a result of the need is to focus on providing suitable abstractions for dealing
for more efficient representation and interpretation of context with heterogeneity and context dissemination. A representative
(e.g. Cooltown project [42]). There may also be multiple example is MobiPADS [48] middleware system, which first
layers that context information goes through before reaching introduced the idea that both centralized network infrastructure
a CA application due to the need for additional abstraction and MT context information have to be taken into account for
and thus multiple middleware implementations (between two a decision to be made. PACE middleware’s vertical handover
layers of abstraction at a time) are required [19]. Next gen- (VHO) use case found in [20], goes one step further describing
eration middleware approaches were even more capable of ways that a CA VHO procedure could be accomplished among
adapting to environment changes and supporting the required heterogeneous coexisting wireless network technologies.
level of Quality of Service (QoS). For example, in CoBrA
[43], SOCAM [44] and CAMPS [45] architectures, ontology Context-dependent autonomic networking approaches:
engineering concepts were adopted in order context inference, Modern CA architectures should provide appropriate mech-
modelling and reasoning functionalities to be upgraded. All the anisms to achieve a suitable balance between user control
pre-referred proposals, including CMF [46] and CASS [47], and CA software functionalities autonomy. That is, CA ap-
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 7
plications not only require middleware for distribution trans- of context modeling (higher-level context) can be exchanged
parency of components, but mechanisms dealing with system among various CAMoWiN heterogeneous entities before it
reconfigurability and adaptation, too. As stated in section serves as input to context evaluation functionality. The various
II, one main cross-cutting issue for CAMoWiN is how CA final CAMoWiN system actions generated as output from var-
autonomic networking can be realized. CONTEXT Project ious CAMoWiN decision making entities can provide strategic
[49] was the very first research initiative trying to introduce business decisions according to human-oriented business logic
ideas on how context-awareness, ontology engineering and principles. By the term ”horizontal functionalities”, we mean
autonomic networking concepts can coexist in an innovative, all CA operations that can be applicable in any other CA
extensible and scalable knowledge platform. Moreover, policy- functionality. For example, security, privacy and trust man-
based network management (PBNM) [50] concepts have been agement mechanisms (see section IX) can be applied in more
exploited in CAMoWiN in the essence that a CA architecture than one CA functionality. Energy efficiency could be another
should support functionalities for sensing context changes and horizontal functionality as well as paradigms related with
using policies specific to the new context in order to provide multi-disciplinary research. Conclusively, figure 3 represents
feedback to decision making entities [23]. Several CA ar- the roadmap of this paper as it sums up the content of
chitectures concerning CAMoWiN support context-dependent the following sections, which is well balanced regarding the
decision making functionalities such as CAPP [51], CA3RM- networking and computing perspectives.
Com [52] AISLE [53], E3 [24] etc.
Latest state-of-the-art CA architectures: Remarkable re- IV. ACQUIRING THE C ONTEXT
search work has been done in designing CA vertical mobility The most common functionality for CAMoWiN field is
management architectures during the last few years. Cross-
the context acquisition. In general, raw data, modulated on
layer and cross-system architectures were designed and im-
any carrier (i.e an electromagnetic wave), are transmitted by
plemented for 4G network systems mainly dealing with CA context sources via the wireless medium and are received
resource management challenges seen from an overall system
by appropriate context acquisition modules. Following all the
perspective (see section II). CARP framework [54] proposed
stages of context acquisition functionality described in this
ways to proactively manage network resources in a 4G het- section, raw data are then transformed into low-level context
erogeneous wireless environment. Tramcar [55] and CrossTalk
information serving as input for the next CA functionality
[21] architectures first introduced the idea of exploiting global to be presented in the next section. At the early days of
networking context information in order to optimally design
deployment of CAMoWiN applications and services, each
and implement various CA functionalities. Working further on
developer used to implement his own context acquisition
this perspective, several CA architectures include IEEE Media architecture preventing reusability and facing vital scalability
Independent Handover (MIH) framework [56] and/or IEEE
problems. As a result, the advent of a general-purpose context
P1900.4 functional architecture [57] such as CAMMS [58],
acquisition framework has become a necessity. Designing such
HURRICANE [22] [59], EMIH [60] and the P1900.4-based a framework inherently involves several challenging issues
architecture proposed in [61], while [62] demonstrates ways
mainly due to its scale and sophisticated sub-functionalities,
that CA systems can benefit from a generic machine learning
which will be further analysed below. As stated in [64], context
framework. The idea that CA can be seen from a holistic
acquisition approaches should satisfy three requirements: a)
interactional point of view (i.e. context viewed by the prism of
easy to deploy, b) easy to use, and c) non-intrusive for the
activity) was first introduced in BAC framework [13]. Finally, end users. Taking all these into consideration and having
up-to-date research trends propose the incorporation of: a)
made a thorough investigation in the international literature,
business logic (e.g FOCALE architecture [23]), b) context-
we propose four main sub-functionalities of a CAMoWiN
driven content creation, adaptation and media delivery via architecture regarding context acquisition: a) monitoring, b)
personalization services provisioning (e.g. C-CAST [6], CA-
gathering, c) predicting, and d) learning the context. In the
SUP [28]) and c) innovative security policy frameworks (e.g.
rest of this section, the four sub-functionalities are further
SemEUsE architecture [63]) into next-generation CAMoWiN
broken into distinct areas (sub-functionalities) and thoroughly
systems.
analysed, following the structure of the taxonomy scheme
As depicted in table II, the various CA architectures and depicted in figure 4.
frameworks are evaluated according to the extent of sup-
port they provide for eight different CA functionalities. We
summed up these functionalities to six and a complete analysis A. Monitoring the Context
is provided for each one of them in sections IV-IX. As In order to provide CA services to mobile users, there must
depicted in figure 3, these CA functionalities are: a) context be specific entities, which have to continuously monitor all
acquisition, b) context modeling, c) context exchange, d) types of contexts and capture their changes over time. A
context evaluation, e) exploitation of context from business monitoring framework should deal with a broad spectrum of
logic perspective and f) a major CA horizontal functionality contexts, from personal to worldwide in terms of scale, and
regarding security, privacy and trust issues. More specifically, from crude to highly processed in terms of complexity. For
the context information flow is as follows: after raw context example, in [65], context is classified in three main layers
data is acquired from context sources and is appropriately regarding personal, regional and global scale while in [66], it
processed, it is then transformed into low-level context, which is clearly stated that a context monitoring framework should
serves as input to context modeling functionality. The outcome have the capability of monitoring physical to application-layer
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, ACCEPTED FOR PUBLICATION
Predict
estimation gathering and pre-processing mechanisms are then consid-
evaluation
auto/manual ered in order to conclude with the final low-level context
feedback information (the outlined area in figure 4). A single mobile
supervised
device or sensor is generally not capable (due to processing
capacity, memory and battery limitations as well as due to
Learn unsupervised
the inherent heterogeneity and crudeness of raw context) to
semi-supervised process the information facing it as a one-step procedure. More
specifically, raw context data have to be aggregated, classified
Fig. 4. Context Acquisition Taxonomy scheme and labelled.
During the aggregation phase, device (e.g. sensor) readings
are mapped into feature values such as mean or variance, using
context information via appropriate cross-layer mechanisms. statistical and signal processing techniques, in order to reduce
In its infancy, monitoring in the CAMoWiN area dealt only computational and communication overhead. For example,
with simple collection of measurements of predetermined instead of having vast sets of measurements, we can have
parameters by network devices (e.g. SNMP MIBs, logs, etc.) a smaller set of their mean values. In addition, aggregating
[67]. Nowadays, it is accepted that such a cumbersome mon- algorithms recognize similar sets of data in order to merge
itoring framework is inefficient for CAMoWiN environments. them and/or eliminate redundancies [71]. After filtering all the
Therefore, CAMoWiN monitoring components should be con- unnecessary information, the classification task is responsible
tinuously adapted to an ever changing network infrastructure to find common patterns in the feature values. These so
following thus self-configuration concepts. Moreover, self- called classes or clusters represent initial semantic states
healing functionalities can be applied by increasing the CA of such as ”high temperature” or ”crowded place” regarding
monitoring information being exchanged (see more in section sensors indicating weather conditions and population density
VI) while self-optimization can be achieved by exploiting CA correspondingly. The labelling task assigns descriptive names
information in order to take decisions according to specified to combinations of classes in order for the low-level context
administrators’ policies (see more in section VII) [66]. information to be better interpreted and modelled during the
next main CA functionality step.
According to figure 4, the first sub-functionality of monitor-
During the entire gathering procedure and especially for
ing is the sensing of context, meaning both the detection and
classification and labelling tasks, the performance evaluation
reception of data from numerous sensors residing in a specified
of the algorithms being implemented is estimated on the basis
CAMoWiN environment without introducing additional traffic
of predictions delivered by trained models because of the
overhead. In contrast to sensing, active monitoring approaches
uncertain nature of context information [72]. In the rest of the
need to be provisioned in a CAMoWiN monitoring framework,
section, we describe how prediction and learning mechanisms
too. Efficient scanning is a matter of vital importance in
can improve the accuracy of low-level context information.
CAMoWiN, since real-time limitations have to be satisfied,
disruptions of running services should be avoided and users’
QoE (Quality of Experience) objectives should be maintained C. Predicting Lower-Level Context
within acceptable levels [68]. Moreover, scanning includes In early CAMoWiN systems, mobile devices were able only
device and service discovery in order for the optimum set to react to the context being acquired. Authors in [73] were
of context sources to be monitored and thus, to maintain among the first who envisioned proactive mobile devices via
an equilibrium between the needed accurate view of the predicted user contexts. Since then, context prediction was
CAMoWiN area and the processing overhead [69]. Following mainly based on high-level context information, because the
this principle, an optimum selection of multi-level contexts has number of higher-level contexts was considered relatively low
to be made during composition phase. For example, a young (read more in section V). However, recent trends assess that
couple decides to enjoy the weekend in a picturesque town in future CAMoWiN systems, complex scenarios are going
without knowing much about its stores, restaurants, culture to emerge and thus the number of higher-level contexts will
etc. Before leaving their home, they set up their schedule increase abruptly. In [74], it is shown that basing the prediction
using their mobile terminal (MT) and during the day, the MT procedure on lower level context can better deal with the
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 9
contexts
uncertainty of context information. As a result, prediction aggregation
and learning sub-functionalities are included in our context Interpret
common-level
interpretation
acquisition taxonomy scheme. categorization
Prediction algorithms, regarding CAMoWiN acquisition
frameworks, are usually applied to the classified context data inferring new
knowledge
and their outputs are of major importance since possible erro- uncertainty
reasoning
neous low-level context will definitely incur bigger problems Reason
in higher context abstraction levels. An overview of the most Final higher-level
context
important prediction algorithms is given in [74]. Confidence Modeling
the context
estimation is a task of major importance for every prediction timeliness
algorithm because the dynamics that govern the changes of Store/Retrieve filtering
lower-level contextual variables are chaotic and thus highly
refreshment
unpredictable. Therefore, each algorithm has to compute an
estimation of the correctness of the forecasted context, which predictions
can be used as metric for the estimation evaluation task as well evaluation
learning
as for higher-level functionalities and decision making entities Predict/Learn
frequency
(read more in sections V and VII). Estimation evaluation is auto/manual
feedback
based on the output of the confidence estimation task and
some analytical procedures regarding lower-level context. In Fig. 5. Context Modeling Taxonomy scheme
[71], the theory of dynamical systems is proposed as a solution
to effectively evaluate future contexts. Finally, a CAMoWiN
prediction framework should support automatic as well as learning uses a large set of unlabelled data accompanied
manual feedback of context information. That is, future con- by labelled data in order to build better classifiers [77].
text has to be compared with real context when this becomes Conclusively, supervised learning algorithms require enough
available in future time series and consequently provide input labelled training data to learn reasonably accurate classifiers
to confidence estimation and estimation evaluation modules. (i.e. time and resource consuming procedures) and thus they
At the manual feedback case, whenever the user cancels are used in cases where non-real time data need to be
abruptly some action that has been carried out automatically processed. Unsupervised learning methods can be employed to
due to a forecast, this fact is taken into account for future discover structure in unlabelled data in cases where quick and
estimation and evaluation tasks [73]. sub-optimal solutions are required. Finally, semi-supervised
learning allows taking advantage of the strengths of both by
employing flexible use case-specific mechanisms according
D. Learning Lower-Level Context to higher-level policies ensuring thus CAMoWiN system’s
In CAMoWiN environments, both learning and adaptation robustness and simultaneously utilizing system resources in
to the users’ behavior and their surrounding environment is the most efficient manner (read more in section VII).
necessary for context prediction, as a decrease in learning
accuracy severely affects the context prediction quality [74]. V. M ODELING THE C ONTEXT
The main challenges of learning lower-level context regarding Final low-level context needs to be further processed in
CAMoWiN systems are a) learning to classify low-level a CAMoWiN system. As already stated in the introductory
sequences (e.g. inferring actions from location and goals section, context modeling is a functionality that interprets low-
from actions), b) learning to segment sequences into discrete level context into higher-level context, reasons it to derive
activities (i.e. learning to classify flows of context), and c) further implications and represents it in ways that efficient
correct mapping between the currently observed time series storage, retrieval and learning mechanisms can be applied.
to the time series that will be observed in the future (i.e. Moreover, this main functionality’s output, which is the final
provide feedback to context prediction sub-functionality). Ma- higher-level context (see the outlined area in figure 5), has
chine learning has extensive applicability in CAMoWiN field to be in a standard format so that both context exchange
and generally refers to the study of algorithms that improve and evaluation functionalities can efficiently exploit it as their
automatically through experience [62]. Machine learning tasks input for their own processes (see more in sections VI and
can be addressed through supervised, unsupervised and semi- VII). Requirements set for context models being adopted by
supervised learning. CAMoWiN systems have been investigated in recent works
Supervised learning is based on training data and includes such as [78], [79] and can be categorized in: a) easy manipu-
two main steps: a) a training step where classifiers learn from lation on many devices supporting heterogeneity and mobility,
training data and b) a prediction step where class labels are b) providing low overhead in keeping models up-to-date, c)
assigned in order to test data [75]. In the case of unsupervised easy extension and reusability of modeling frameworks, d)
learning, no labelled data are available for training and thus efficient uncertainty management, e) timeliness, f) semantic
only sub-global classification is computed. In CAMoWiN reasoning, and g) scalability.
systems, unsupervised learning can have remarkable appli- Early approaches had limited capabilities in satisfactorily
cability due to the fact that external supervision is avoided addressing many of the above mentioned prerequisites. Key-
during context acquisition phase [76]. Finally, semi-supervised value, mark-up scheme, graphical and logic-based models are
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
10 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, ACCEPTED FOR PUBLICATION
some of these early approaches presented in [69], whose context information [79]. In CAMoWiN systems, three types
main drawback is that the context model is tightly coupled of reasoning mechanisms have to be implemented, namely, a)
to the rest of the CAMoWiN system. Nowadays, the most rule-based, b) ontological and c) probabilistic [83], towards
widely adopted context modeling approaches are: 1) object- achieving the ultimate goal which is to find the optimal trade-
role, 2) spatial and 3) ontological. Ontologies are the most off between expressiveness and complexity of reasoning. In
expressive context representation models but they have poor table I, we consider context to be measured and inferred
computational performance during reasoning process. There- knowledge instead of just a set of numerical values. Ontologies
fore, hybrid and hierarchical hybrid context models seem provide a formal specification of contexts’ semantics and are
to be the best solution, by combining advantages of two thus well suited for knowledge sharing (see more about context
modeling approaches for fast responsiveness of CA services exchanging in section Context Evaluation Taxonomy scheme
in CAMoWiN systems [78], [80], [81], [82]. At the remainder VI. During the last decade, ontological reasoning has been
of this section, we present four main sub-functionalities of a applied in several CAMoWiN architectures such as [23], [28],
CAMoWiN architecture regarding context modeling and we [43], [44], [51], [84]. Nevertheless, it incurs significant compu-
analyze them following the structure of the corresponding tation overhead, which is critical for real-time CA applications
taxonomy scheme depicted in figure 5. and hence hybrid knowledge representation formalisms and
loosely coupled reasoning procedures are adopted in recent
A. Interpreting the Context works [78], [80].
Context reasoning sub-functionality consists of two main
In order to obtain high-level context information, the fi-
tasks, namely a) new knowledge inference and b) uncertainty
nal low-level contexts (i.e. the output of context acquisition
reasoning. Regarding the former one, new context can be in-
functionality) have to be further processed in the context
ferred based on information retrieved by databases (see section
interpretation step. In the absence of this sub-functionality,
V(c)) and new raw context data acquired by context sources
low-level context data can be meaningless, trivial, vulnerable
(see section IV). This task is put in practice by inference en-
to small changes and uncertain. All these can directly affect the
gines, which implement rule-based and probabilistic reasoning
usefulness and soundness of CA applications and hence good
mechanisms in conjunction with pre-referred ontological ones
interpretation of low-level contexts becomes indispensable.
[19]. Reasoning on uncertainty substantially means improving
Early implementations had included interpretation as a sepa-
the quality of context information. This improvement can be
rate functionality [34], [35]. However, even nowadays, the task
achieved by correctly recognizing and classifying the four
of rightfully interpreting huge amounts of context information
types of uncertain context information described in section II
is yet a challenge due to its uncertainty and superfluity [46].
(figure 1). For example, it is of major importance to distinguish
According to figure 5, the first task of the interpretation phase
between wrong and imperfect as well as between wrong and
is the contexts aggregation. Low-level context data ”carry”
unknown information [16]. Detection of possible inconsis-
a huge amount of information that has to be filtered in
tencies or conflicts falls also within uncertainty reasoning
order to be manageable by a CAMoWiN system regarding
task’s objectives. Some of the most common approaches used
modeling functionality. Therefore, aggregation and fusion of
for this task are: fuzzy logic, probabilistic logic, Bayesian
related contexts, data calibration, noise removal and reforming
networks, Hidden Markov models, and the Dempster Shafer
of data distributions are some sub-tasks that are applicable
theory of evidence [78]. Concluding this task, we have to
[74]. Another main objective of context interpretation sub-
outline that reasoning tasks cooperate with the other modeling
functionality is to provide a common level of interpreted
sub-functionalities such as prediction/learning tasks for better
context going to be reasoned. This is not a trivial task as
inference outcomes as well as with storage/retrieval tasks for
context information models deal with a large variety of context
better uncertainty reasoning. The final outcome of modeling
sources that differ in their update rate and their semantic
functionality is the higher-level context outlined in figure
level [78]. For example, context obtained from a CAMoWiN
5, which will serve as input for exchanging and evaluation
system’s distributed databases is often heterogeneous and thus
functionalities presented in the next sections of this work.
a common level of interpretation has to be applied in order
system’s consistency and robustness to be achieved. Finally,
an additional categorization of context is required whose C. Storing and Retrieving the Context
outcome will serve as input to reasoning sub-functionality.
In CAMoWiN systems, persistence is called the process of
Hence, the interpreted contexts can be categorized in sets each
storing context in databases, in ways that semantics afforded
one of them providing identification, activity, spatial, temporal
by the context model are maintained and efficient retrieval
information etc [2].
is facilitated. This means that storage and retrieval function-
alities are tightly coupled to context reasoning mechanisms
B. Reasoning the Context [69]. Generally, storing all context information in CAMoWiN
Context modeling can be applied in facts that are: a) systems is impossible and technically unacceptable in terms
observed or measured and b) inferred. In the former case, of retrieval efficiency [46]. In [73], it is stated that the higher
reasoning process can be eliminated whereas in the latter one, the level of the recorded data, the less data typically needs
facts that are inferred can only exist by using reasoning to to be stored, but the higher the effort for retrieval. Therefore,
generate them [14]. Hence, reasoning refers to information that it is widely accepted that storage and retrieval mechanisms
can be inferred from analyzing data and combining different worth investigation in terms of modeling functionality even if
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 11
important are: a) scalability, b) interoperability, c) adaptabil- Context consumers are the final receivers, users and/or
ity, d) dealing with context uncertainties, e) content-based evaluators of the context information being exchanged [99].
dissemination, f) application-independent context exchange, In a future CAMoWiN environment, it will be common for
and g) light-weight protocols’ implementations. Privacy and an entity to act as context source for a CA application at
security is another basic requirement but it will be extensively a specific time interval and as context consumer for another
presented in section IX. Regarding all the other prerequisites, CA application at a subsequent time interval. For example, a
at the remainder of this section, plethora of research proposals mobile terminal’s user can enjoy a CA service offered by a
published in the past will be reviewed, state-of-the-art issues shopping mall (e.g. acquire and use a list of shops she would
will be presented and emerging open issues and challenges like to visit acting thus as a context consumer entity) as well as
will be discussed. According to figure’s 6 structure, the main provide context information via a social network site regarding
CA functionality is split in three sub-functionalities as we want the shops she has already visited for other people entering the
to a) define and categorize all possible entities (i.e. ”who”) place (acting thus as a context source entity). Given the basic
that are participating in the emerging CAMoWiN ecosystem provider-consumer model, only specific-purpose CAMoWiN
[95], b) define the parameters used for deciding ”what” has to systems could work with direct communication between con-
be exchanged at any time instance among the various actors text sources and consumers. However, in future large-scale
discussed at (a), and c) define and categorize the means (i.e. CAMoWiN environments, the existence of some kind of inter-
”how”) by which efficient end-to-end CA communication can mediaries or the so called context brokers becomes a necessity
be achieved. in order transactions costs to be reduced as well as scalability
and interoperability challenges to become manageable. Several
recent works deal with broker-based models. For example,
A. Actors [94] describes a peer-to-peer infrastructure for context distri-
As actors, we mean all the functional entities of a bution, where disseminator peers act as the most important
CAMoWiN system, which are interacting with each other in components of the proposed context exchanging framework.
order to exchange context information and thus be able to In [95], business logic proposals are discussed regarding the
make the optimal decisions (see section VII). As we have roles context broker entities will play in future CAMoWiN
already stated in section III, early CAMoWiN systems adopted ecosystems, while in [99], a federated broker architecture and
centralized approaches in their architectural design developing means of inter-broker communication are proposed. Finally,
small-scale systems for specific-purpose CA applications (e.g. [100] proposes ways of controlling context quality (i.e. CQE
Context Toolkit, CoBrA, Cooltown, etc.). Consequently, the component residing in context broker entities) according to
greater the scale of CAMoWiN systems, the more decentral- the prerequisites defined by context consumers. Despite the
ized approaches were adopted as the years were going by recent research being done so far, there is plethora of open
(e.g. PACE, MobiPADS, CA3RM-Com etc.). Nevertheless, issues regarding the roles all the pre-referred actors will play
the main drawback of these decentralized approaches is the in future CAMoWiN real-market scenarios.
fact that they are lacking of a set of common commu-
nication protocols, which would define standard means of
exchanging context information among CAMoWiN entities. B. Content
P1900.4-based [61] [96], HURRICANE [22] and CAMMS Context-based content delivery is a research challenge,
[58] are some representative example CA architectures, which which deals with ways to provide personalized content to
paved the way for meeting scalability and interoperability context consumers taking into account all contextual hetero-
challenges by adopting IEEE P1900.4 [97] and IEEE 802.21 geneities of a CAMoWiN environment. Multiparty delivery
[98] standards. concepts have already been standardized and developed in
State-of-the-art research dealing with CAMoWiN actors mobile and wireless systems using multicasting and broad-
[94] [95] [99] [100], categorize them in three main sets named casting technologies to deliver the same content to specific
context sources, context consumers and context intermediaries. groups of users [88] [89]. However, these technologies should
As context sources, we mean all CAMoWiN entities, which be enhanced in order multiparty session creation, management
can generate and provide context information. Nowadays, we and autonomic operation to become a reality [101]. Ultimately,
can classify context sources in three main categories [95]: a) exchanging context functionality has to be content-based in-
location information providers, b) social network sites, and stead of channel or subject-based [87].
c) wireless sensor networks. Based on the source of context In early CAMoWiN implementations, context exchanging
information, sensor-equipped mobile terminals, telecom oper- was performed following specific mechanisms and certainly
ators, sensor network operators, Internet platform operators, adopting minor CA concepts [93]. However, state-of-the-
ICT companies, individual middleware developers etc. can be art CAMoWiN architectures require good quality of context
classified in one or more of the pre-referred categories. Based (QoC) to be exchanged to achieve efficient content delivery
on the two main CA functionalities described at the previous (regarding overall system’s perspective performance evalu-
two sections, a context source deals with context acquisition ation metrics) [102]. More specifically, the contextual un-
and (sometimes) context modelling (depending on its type certainties demonstrated in figure 1 should be manageable
and capabilities as well as the type of the CA application) and the corresponding uncertainties’ thresholds should be
before the context information is delivered to other CAMoWiN dynamically defined at regular time intervals for each CA
entities. application. Therefore, CA features will be exploited only if
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 13
derivation/
specific QoC constraints are satisfied in an end-to-end CA enforcement
communication framework [69]. Given that these constraints efficiency
Policies
evaluation
are satisfied, user/group-oriented content delivery should be conflicts
resolution
applied. That is, although the same content is to be sent
to all group members, its delivery should be adapted for resource
management
different sub-groups based on their specific context. Efficient
Decision user satisfaction
sub-grouping mechanisms are proposed in works such as Evaluating Making
[93], [101], [102] and [103], which use cross-layer contextual the context Final system
actions
information (e.g. user’s profile, mobile terminal capabilities,
network congestion, available bandwidth, link quality and configuration
C. Protocols
Conclusively, many open challenges are related with scalabil-
Having described all possible participants (i.e. actors) of ity and interoperability issues and CAMoWiN area’s future
a CAMoWiN ecosystem and the type of context informa- success depends on their effective troubleshooting.
tion (i.e. content) to exchange, here we focus on efficient
ways to exchange the content among the various actors. CA VII. E VALUATING THE C ONTEXT
session management frameworks should be flexible enough
Summarizing the previous three CA functionalities, we have
and support several tasks including session initiation, modifi-
cation, termination, dynamic sub-group adaptations, session come to a state where all actors of a CAMoWiN system
autonomic operation etc [93]. For example, a sequence of possess all the higher-level context information they need in
order to proceed with their final decisions and actions (see
signalling messages for context-based content delivery can
differ in situations where a) strict QoS constraints have to outlined area in figure 7). According to [46] [69], context
be satisfied, b) energy consumption has to be minimized, c) evaluation functionality requires semantic richness of context
information as input in order a CAMoWiN system to be
specific security and privacy levels have to be achieved and
d) unnecessary data transmission and processing in terms of able to change its behaviour in response to a context change.
Past network management approaches lacked the ability for
optimal system performance can be avoided [104]. Therefore,
by knowing and predicting the context, plenty of delivery dynamic context adaptability, suffered from time-consuming
options can be available regarding quantitative, qualitative and and low-quality decision-making procedures while general
systems’ performance was strongly constrained by humans’
sequential features of messages being exchanged and thus
optimal system performance evaluation achievements can be expertise [66]. In today’s CAMoWiN environments where a
realized (read more in section VII). vast number of cognitive mobile terminals, innovative mobile
applications and different types of access networks are avail-
By scalability term, we mean ways that context exchange
able, evaluation should be context-dependent and consider sev-
protocols can deal with continuously growing factors such as
eral factors, at different abstraction layers [25]. More specif-
a) number of actors, b) number of interactions in terms of
ically, a context evaluation scheme should at least satisfy the
message exchanging between the actors, c) area of interaction
following general prerequisites: a) efficiently handle multiple,
and d) time span (more context has to be available for longer
dynamically changing and potentially unexpected situations
periods) amongst others [65] [103]. In [99], various techniques
[69], b) guarantee optimum system’s resources management,
are proposed for removing explicit dependencies between in-
c) keep users’ satisfaction levels above predefined thresholds
teracting actors facilitating the design of scalable CAMoWiN
providing personalized services [106], d) deal with complex
systems. [101] investigates the ”scalability vs. personalization
optimization problems supporting real-time decision-making
trade-off” problem, suggesting optimal solutions that provide
procedures [107] and e) minimize human intervention [108].
personalized CA services without downgrading scalability
At the remainder of this section, analyzing the context
milestones. Research challenges related to interoperability
evaluation functionality, there are policies according to which
emerge because of the large scale of CAMoWiN environments,
the decision making procedures are applied. Self-x properties
too. HURRICANE architecture [22] [105] implements IEEE
can provide feedback information to decision making entities
802.21 MIH protocol [98] in order to provide interopera-
in order a CAMoWiN system to be able to learn from its
ble means of communication among heterogeneous wireless
activities towards finding optimum solutions that fulfil the
network technologies while P1900.4-based architecture [61]
aforementioned prerequisites (figure 7).
[96] provides interoperable means of communication among
abstract mobile terminal and core network entities. [100] pro-
poses an over-the-top implementation of 802.21 MIH protocol A. Policies
in order interoperability to be realized in mobile cloud com- While context conceptualizes actors, content and exchange
puting paradigms, while [95] investigates possible business protocols (see section VI), policies are used for abstracting
roles that actors will have in future CAMoWiN ecosystems. specific actors’ objectives at various levels such as business,
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
14 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, ACCEPTED FOR PUBLICATION
technical and administrative. Policies can be applied to sets of CAMoWiN systems whose scale has rapidly grown during
resources to configure them uniformly and adapt their behav- the last years. The larger the number of policies and the more
ior to suite changing requirements, while possible policies’ complex they are, the greater the likelihood of a conflict and
amendments lead to changes in goals (instead of specific this explains this task’s ever increasing importance. Policies’
system configurations) that are implementation independent conflicts importance degree can be estimated based on: a) the
[108]. There are three main abstract types of policies. Action policy types involved and the scope of their enforcement (e.g.
policies are based on event-condition-action rules whereas resource management policies may have greater impact factor
goal policies specify a desired state and utility policies ex- compared to scanning policies), b) application environment
press a value for each state to indicate how desirable it constraints (e.g. a radio access network (RAN) selection policy
is [109]. Since ’90s when policy-based management (PBM) may opt for a RAN that is not available at a given time
concepts (regarding mobile and wireless networking) came in instance), c) nature of their occurrence, and d) the time frame
the research foreground [110] [111], almost all of the initial at which they can be detected (static vs. run-time conflicts).
proposals or ideas are being adopted in today’s CAMoWiN Especially for CAMoWiN systems, run-time conflicts efficient
systems, too. However, despite the good feasibility and de- handling is a challenge of major importance as the smooth
sign studies having been done so far, PBM’s applicability operation of delay-sensitive CA applications is imperative.
in real CAMoWiN systems still remains below expectations. [15] provides a thorough discussion of the main notions of
This is due to the inherent uncertain characteristics of the policies conflicts resolution task based on the identify-classify-
continuously evolving CAMoWiN field described at previous detect-resolve principle.
sections of this work. Recent works focus only in a few
types of policies at a time, namely: a) generic resource
management policies [61], b) radio access network selection B. Decision Making
policies [112], c) generic handover policies [105], d) dynamic Decision-making frameworks and algorithms were usually
spectrum assignment policies [107], e) self-organizing net- not context-aware and did not take into consideration context
working policies [24], f) scanning policies [68], g) energy- information from several sources simultaneously. [115] is one
saving policies [113] and h) security policies [114]. We assess of the earliest works dealing with resource management issues
that future policy-based CAMoWiN evaluation frameworks in mobile and wireless networking. In 00’s, when many new
should simultaneously take into consideration many of those wireless network technologies had been standardized, novel
types of policies. Regardless of the applied type of policy, challenges came up such as a) heterogeneous environment,
there are certain tasks that have to be supported, namely: b) multiple types of services, c) adaptive and cognitive re-
a) policies derivation and enforcement, b) policies efficiency sources allocation, and d) cross-layer design exploiting context
evaluation and c) policies conflicts resolution. information from multiple abstraction layers [116]. Autonomic
All these tasks run mainly in entities at the network side, decision-making is another cross-cutting challenge (see figure
which have the capability to collect CA information on all 2) and refers to the ability of a CAMoWiN system to perform
mobile terminals (MTs) of a CAMoWiN system. The aim of adaptation operations using its internal knowledge to decide
a policy derivation module is to derive policies, whose goal why, when, where and how adaptations are performed, with-
is to provide MTs with guidelines towards making optimum out any human intervention in the decision making process
decisions in order to ensure profitable operation for both [66]. [93] studies efficient ways to make decisions about
networks and MTs. The enforcement of the derived policies to delivering context information to specific context consumers’
each MT highly depends on the policies’ grade of obligation subgroups in order multiparty multicast session management
(GoO); that is the degree of importance that each policy has procedures to be realized. [100] proposes ways on making
according to its: a) purpose, b) imminence, c) impact, d) intelligent decisions regarding network resources management
freshness, e) complexity etc [96]. Generally, since policies in mobile cloud computing environments. Vertical handover
are exchanged via unreliable wireless interfaces, they can (VHO) decision-making is also a ”hot” topic in CAMoWiN
be seen as high-level context information that brings on all field, while possible synergies between IEEE 802.21 [98]
types of uncertainties described at figure 1. Hence, MTs’ and IEEE P1900.4 [97] standards seem to provide good
decisions may not be the desirable ones and out of the initial architectural solutions [58] [105] [117]. Finally, works such as
scope of their enforcement. Policy efficiency evaluation task [118] and [96] investigate the distribution of decision-making
aims at enhancing the knowledge acquisition capability of functionality between MTs and the network side trying to
the policy derivation task and thus improving the efficiency define performance indication keys regarding users satisfaction
of future policy derivations [105]. In a nutshell, this task levels.
should be cognitive enough to ”understand” whether bad value Based on the literature and according to figure 7, we can
indications happen due to: a) bad wireless channel conditions, categorize decision-making solutions as: a) generic resource
b) bad information handling and decision making by MTs or management mainly accomplishing operators’ objectives and
c) the fact that current policies in deed have to change (out- b) personalized decision-making for each individual MT
of-date, conflicting, not applicable any more, etc.) in order mainly satisfying end user needs. Till recently, there was not a
the efficiency evaluation reports back to the policy derivation clear difference between the two pre-referred decision-making
modules to have a high degree of effectiveness [61]. Finally, variants. For example, our proposed call admission control and
conflicts resolution task aims at detecting and resolving ex- scheduling algorithms [112] [119] try to keep some impor-
isting and potential policies’ conflicts and inconsistencies in tant metrics (delay, blocking/dropping probabilities, number
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 15
input from users b) no sufficient users’ history and long-term interactions among humans. In [133], it is stated that it is
context information regarding users’ activities, c) no efficient very doubtful that the society is adequately equipped with the
ways of retrieving users’ preferences, d) little consideration in right sociological and technical tools in order to understand
framework and architectural design of the users’ preferences what can happen when everything is available, knowable,
to recommend the personalized services on CA applications, searchable and recorded by everyone all the time (e.g. in social
etc [28] [77]. networking sites). From context sources perspective, it is very
Referring to personalization sub-functionality in today’s important to take into consideration the socio-cultural profile
CAMoWiN systems, we argue that it can be split into three of context consumers in order to avoid useless delivery of
main tasks, namely a) profiling of consumers, b) delivery of context information that may even frustrate end users (for
selective services and c) evaluation of appropriate impacts. example imagine the case of a regular employee getting a
More specifically, a complete personalization component has message about a special trendy cloth discount costing ”only”
to collect comprehensive context information and somehow 1000 euros!). Psychological aspects have to also be taken
interpret it into actionable knowledge stored in the form of into account, since a wrong decision-making procedure by
consumer profiles. Thus, it is of major importance to find a CAMoWiN system can irritate end users at the extent of
ways to construct accurate consumer profiles by utilizing abruptly stopping the usage of a specific CA application for
advanced profiling techniques in order to capture semantic- a long time. Minimizing intrusiveness is also important, since
rich behavioural information [128]. Based on the knowledge CAMoWiN systems should not give the impression to users of
about each mobile end user stored in the consumer profiles, the an outsider always monitoring them. According to [128], it is
personalization component has to match appropriate content challenging to find optimal trade-offs between the usefulness
and services to individual users or groups. Several matchmak- of recommendations and the amount of privacy the consumer
ing and content-delivery methods are proposed in [101] [128] needs to give up to receive acceptable-quality personalized of-
[129] [130], trying to achieve optimal solutions regarding ferings as well as determining ways to minimize intrusiveness
the personalization vs. scalability trade-off problem. Finally, (at a data acquisition phase - see section IV) while maintaining
personalization impacts evaluation task determines the extent personalization quality (at a context evaluation phase - section
of users’ satisfaction regarding the delivered personalized VII). Finally, CAMoWiN systems should extend the human
services. This kind of information can act as feedback for subconscious functionality by acting at the right time and
possible improvements regarding the operations of the two by providing the right level of information. Simultaneously,
previous tasks. Conclusively, value-creating elements of per- all these can happen only if ethical and legal issues are by
sonalization are evaluated and thus enhanced personalization no means compromised. Conclusively, ICT and human values
policies and rules can be derived in order better personalization have to be integrated in such ways that the former enhances
decisions to be made [124]. and protects the latter, rather than damaging them [134].
CAMoWiN systems, the use of context information during Another hot issue in the CAMoWiN market is the eco-
policy definition and access control enforcement improves nomics of privacy, which studies the trade-offs associated
adaptability and flexibility, stemming several proposals for with the protection or revelation of personal information.
Role-Based Access Control (RBAC) models [138]. From the economical point of view, privacy is a state of
asymmetric distribution of personal information among market
participants [124]. On the one hand, a big question is what
B. Privacy would be the monetary incentives for users in order to disclose
Privacy is the claim of individuals, groups and institutions their personal data at a given situation. On the other hand,
to determine for themselves, when, how and to what extent users may soon have to pay for privacy (privacy paradox),
information about them is communicated to others [141]. as publicity is becoming increasingly free. Solutions to these
There are six main requirements for a CAMoWiN system to challenges depend on the individual perception and evaluation
support privacy [143]: a) notice (collection practices, privacy of privacy and thus multi-disciplinary research is required
policies and policy announcements should be efficiently de- (read more in sections VIII(C), VIII(D) and IX(C)). How-
clared), b) choice and consent (users should have the choice ever, apart from technical and financial perspectives, privacy
of carrying out, or not, their personal data), c) anonymity is primarily a human right and thus corresponding global
and pseudonymity (whenever the users’ identity is not re- law protection frameworks should be addressed. Nowadays,
quired or whenever the user does not consent, anonymity or the most complete and CAMoWiN-related law framework is
pseudonymity services should be provided), d) proximity and the EU Directive 2002/58/EC concerning the processing of
locality (the collection of data from a user’s device should personal data and the protection of privacy in the electronic
only occur when the user is present while processing and communications sector (continuation of Directive 95/46/EC).
access to these data should only be done within the space of After all, a good law framework being adopted by the majority
collection), e) access and recourse (access to the user’s data of governments and institutions all over the world will assist
should only be allowed to authorized persons) and f) adequate CAMoWiN’s area sustainability because context consumers
security (in terms of the security principles analyzed above). are generally more open to disclosure when legal requirements
Privacy control is essential to be integrated in the design of are respected [140].
a CAMoWiN system, since end users may be unwilling to
use services if the risks of privacy violation, when using a C. Trust
CA application, cannot be estimated. Therefore, appropriate As stated above, privacy of personal data should be sup-
user interfaces for mobile devices enabling users to configure ported by trust management architectures, too. In general, trust
privacy settings as well as privacy advisors, which inform is another mechanism to cope with uncertainty issues and
users of non-obvious possible consequences of their actions consequently it is highly context dependent [139] [145]. More
regarding their privacy, have to be developed [142]. specifically, trust is contingent on time, history, place and
In past CAMoWiN implementations, acts associated with situation, culture, roles, psychology, reputation, recommenda-
privacy protection mainly dealt only with problems between tion etc. In CAMoWiN scope, trust refers to the reliability of
two parties. Nowadays, third parties maintain, access and the context information being delivered and can be evaluated
exploit context information on behalf of other CAMoWiN by computing the distance between it and the real context.
entities (e.g. Mobile Cloud service providers, MVNOs, etc.) As a result, the term ”trustworthiness” is used to represent
and this reality makes privacy problems much more compli- the level of trust that can be assigned to one party (B) by
cated. Generally speaking, privacy of personal data should be another party (A) to do something (X) in a given relational
supported by identity management (i.e. before users release context and thus is a measure of the objective probability that
data) and trust management (i.e. after users release data) the trustees will behave as expected by the corresponding
[140]. Specifically, identity management frameworks enable trustors [136]. Privacy-aware CAMoWiN architectures are
users to limit the amount of personal information revealed to usually proposed and developed without taking into account
certain providers or in certain applications. All CAMoWiN CAMoWiN actors’ trustworthiness [143]. Initial approaches
entities can choose whether they want to identify themselves from the literature assess that just incorporating the human
(by real name or pseudonyms) or remain anonymous at any notion of trust in CAMoWiN trust management architectures
time instance [141]. Anonymity refers to the absence of comprises an adequate solution [146] [131]. However, nowa-
identifying information associated with a natural person where days, it is well accepted that coordinated multi-disciplinary
no claims allowing identification are provided, although other solutions should be adopted. In [139], it is comprehensively
claims might be needed (e.g. non-repudiation). On the other explained how trust is defined in various disciplines such
hand, pseudonymity is the situation, where certain claims are as sociology, economics, philosophy, psychology, organiza-
provided but these cannot be connected to directly obtain tional management and autonomic computing. Consequently,
identification; however, the natural person is still identifiable, the main CAMoWiN trust management requirements can be
if necessary [144]. From the above, it is clear that different summed up to: a) trust should be established based on potential
users can have varying levels of anonymity sensitivity and risks, b) trust should be context-dependent, c) trust should be
different applications can require varying anonymity levels based on each actor’s own interest, d) trust should exploit
for the same user; hence, policy-based dynamic anonymity learning procedures, e) trust may represent system reliability
architectures are required in order to deal with all these and f) efficient management of privacy and security issues
challenges. should be addressed.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
20 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, ACCEPTED FOR PUBLICATION
TABLE III
M APPING OF MCC STRENGTHS AND OPPORTUNITIES TO CAM OW I N OPEN CHALLENGES
Trust management is inherently related to risk management CAMoWiN descendant as the majority of its open challenges
theory because the most important part of managing trust is can be perfectly matched with emerging MCC concepts
understanding the risks involved in trust-based interactions (cf. table III). The main future challenges for the different
[147]. For CAMoWiN field, dynamic evaluation of trust levels CAMoWiN functionalities, which provide grounds for further
in context information itself, is a very challenging issue. research by both academy and industry, are the following:
According to this rationale, CA mobile applications can accept Context acquisition & modeling: a) adaptive software sup-
the risk of using untrusted context information either if they porting context data acquisition from a continuously increas-
are in urge or they don’t really care of the reliability of the ing number of heterogeneous sensors and efficient integration
context information. On the other hand, they can decide to with mainstream business platforms, b) further machine learn-
fix a threshold of trust, below which context information is ing research on distributed robust models learning, data and
not taken into account when considering adaptation of SPT model migration, semi-supervised classification on time series
functionality [26]. Conclusively, subjective and objective prob- etc., c) hardware sensor-like developments (embedded in MTs)
abilities of trust levels (i.e. trust vs. trustworthiness) usually for efficient and non-intrusive users’ activity understanding, d)
differ and this directly results in incorrect risk estimation processing performance and energy efficiency enhancements,
and management. Therefore, careful risk estimation is closely e) further trade-off investigation about lower vs. higher-level
linked with building reliable trust relations among CAMoWiN context information prediction, learning, storage, retrieval,
actors and is one of the most important open challenges for etc., f) integration of diverse reasoning techniques, and g)
the whole SPT functionality. multi-disciplinary research on how context is gathered, stored,
retrieved, inferred etc.
X. O PEN C AMOWIN C HALLENGES & C ONCLUDING Context exchange & evaluation: a) standardization ini-
R EMARKS tiatives dealing with scalability and interoperability issues
Summarizing our work, after defining CAMoWiN area and in a future context ecosystem, b) greater contextualization
presenting the corresponding related work in the introductory of communication services supporting both context-triggered
sections, we analyzed six main CA functionalities (i.e. sections actions and context-dependent reactions [155], c) optimal
IV-IX). In this section, we aim at briefly outlining related trade-off achievements between evaluation effectiveness and
open research challenges providing thus hints for envisioning expressive power of context representations [25], d) relay
the long-term future of CAMoWiN. As a result, we believe and mesh-enabled networks development with cooperative
that Mobile Cloud Computing (MCC) paradigm [100] [125] communication features in OFDMA-based wireless network
[148] [149] [150] [151] [151] [153] [154] can become the technologies [126] [156] and exploitation of spectrum white
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 21
spaces via cognitive radio techniques [157], e) dynamic been investigated in this paper, are expected to encourage
service composition mechanisms provided by novel internet even non-ICT researchers to see ”bigger pictures” of their
architectures for better mobility support and service delivery profession, in which CAMoWiN can play a complementary
[158] [159] [160], and f) energy efficiency issues from MTs, role of major importance paving thus the way for more
base stations and core networks perspectives [161]. targeted future multi-disciplinary research collaborations.
Business logic: a) plethora of new application classes can
benefit from CAMoWiN systems in smart spaces such as smart R EFERENCES
cities, factories or hospitals and other smart environments [1] W. N. Schilit, ”A system architecture for context-aware mobile com-
where (among others) emergency management, remote mon- puting,” PhD Thesis, Columbia University, New York, 1995.
[2] N. Malik, U. Mahmud and Y. Javed, ”Future Challenges in Context-
itoring, vehicular communications and gaming infrastructures Aware Computing”, IADIS International Conference WWW/Internet,
can be deployed [79], b) achieve the virtuous personalization 2007.
cycle and avoid falling into depersonalization traps [128], c) [3] P. D. Reyes, J. Favela and J. C. Castillo, ”Uncertainty Management
in Context-Aware Applications: Increasing Usability and User Trust”,
internet of services and cloud computing visions enhancement Springer Wireless Personal Communications, vol. 56, no. 1, pp. 37-53,
by allowing services to be dynamically composed and enabled 2011.
by technologies for seamless integration of real and virtual [4] S. K. Das and C. Rose, ”Coping with Uncertainty in Mobile Wireless
Networks”, in Proc. 15th IEEE International Symposium on Personal
worlds [6] [131] [158], and d) multi-disciplinary research Indoor and Mobile Radio Communications, pp. 103-108, PIMRC 2004.
as a prerequisite for acceptable CAMoWiN business logic [5] M. A. Razzaque, S. Dobson and P. Nixon, ”Cross-Layer Architectures
solutions. for Autonomic Communications”, Springer Network and Systems Man-
agement, vol. 15, no. 1, pp. 13-27, 2007.
SPT & other horizontal functionalities: a) trade-offs [6] Nigel Baker et al., ”Context-Aware Systems and Implications for
between using context and keeping privacy, b) realize the Future Internet”, in Proc. Future Internet Assembly Book named
best combination of multi-disciplinary trust components to ”Towards the Future Internet - A European Reasearch Perspective”,
edited by G. Tselentis et al., pp. 335-344, 2009.
construct composite trust metrics [139], and c) accountability [7] P. J. Brown, J. D. Bovey and X. Chen, ”Context-aware applications:
and auditability concepts as open issues for trust manage- From the laboratory to the marketplace”, IEEE Pers. Commun., vol. 4,
ment in CAMoWiN systems and Software-as-a-Service (SaaS) no. 5, pp. 58-64, 1997.
[8] R. Hull, P. Neaves and J. Bedford-Roberts, ”Towards situated comput-
cloud models [162]. In addition to all these, energy efficiency ing”, in Proc. 1st International Symposium on Wearable Computers
(EE) can be considered as another CAMoWiN horizontal (ISWC), Cambridge, 1997, pp. 146-153.
functionality whose primary challenge can be summed up [9] D. Franklin and J. Flaschbart, ”All Gadget and No Representation
Makes Jack a Dull Environment”, in Proc. AAAI Spring Symposium
to the fact that optimum trade-off between communication on Intelligent Environments, Palo Alto, 1998, pp. 155-160.
and computing energy should be achieved [161]. Contrary [10] A. K. Dey, ”Understanding and Using Context”, Personal and Ubiq-
to SPT, EE functionality has not been extensively analyzed uitous Computing, vol. 5, no. 1, pp. 4-7, 2001.
[11] J. Strassner and D. O’Sullivan, ”Knowledge Management for Context-
in a separate section because only initial research efforts Aware Policy-based Ubiquitous Computing Systems”, in Proc. 6th
have been made so far. However, we believe that EE-related International Workshop on Managing Ubiquitous Communications and
open challenges accompanied by inter-disciplinary research Services, Spain, 2009, pp. 67-76.
[12] K. Henricksen, ”A Framework for Context-Aware Pervasive Comput-
and technology development will be the cornerstones of future ing Applications”, PhD thesis, School of Information Technology and
CAMoWiN systems and thus this survey paper can be uti- Electrical Engineering, The University of Queensland, Australia, 2003.
lized as a roadmap towards accomplishing future CAMoWiN- [13] R. Sterritt, M. Mulvenna and A. Lawrynowicz, ”Dynamic and Con-
textualised Behavioural Knowledge in Autonomic Communications”,
related research milestones. Springer Verlag Berlin Heidelberg, LCNS 3457, pp. 217-228, 2005.
In table III, the paper ends up by providing a list of [14] J. Strassner, S. V. Meer, D. O’Sullivan and S. Dobson, ”The Use of
convergence points in order to strengthen our claims that the Context-Aware Policies and Ontologies to Facilitate Business-Aware
Network Management”, Journal of Network and Systems Management,
future of CAMoWiN lies in Mobile Cloud Computing. More vol. 17, no. 3, pp. 255-284, 2009.
specifically, some major MCC’s strengths and opportunities [15] M. Charalambides, G. Pavlou, P. Flegkas, J. R. Loyola, A. K. Bandara,
are directly mapped to the CAMoWiN open challenges being E. C. Lupu, A. Russo, N. Dulay and M. Sloman, ”Policy Conflict
Analysis for DiffServ Quality of Service Management”, IEEE Trans.
presented above. For each strength/opportunity, some indica- Network Service Management, vol. 6, no. 1, pp. 15-30, 2009.
tive references from the very recent international literature are [16] K. Henricksen and J. Indulska, ”Modelling and Using Imperfect Con-
provided. At the right column, the corresponding CAMoWiN text Information”, in Proc. 2nd IEEE Annual Conference on Pervasive
Computing and Communications Workshops (PERCOMW), 2004, pp.
functionalities indicate the field in which the various MCC 33-37.
concepts can be applicable. Conclusively, as long as the [17] S. Dobson, S. Denazis, A. Fernandez, D. Gaiti, E. Gelenbe, F. Mas-
integration of computing and networking environments is sacci, P. Nixon, F. Saffre, N. Schmidt and F. Zambonelli, ”A Survey of
Autonomic Communications”, ACM Trans. Autonomous and Adaptive
boosted with the assistance of innovative ICT architectures Systems, vol. 1, no. 2, pp. 223-259, 2006.
and technologies, implications of Cloud Computing paradigm, [18] N. Agoulmine, S. Balasubramaniam, D. Botvitch, J. Strassner, E.
which are applicable in CAMoWiN area will be continuously Lehtihet and W. Donelly, ”Challenges for Autonomic Network Man-
agement”, in Proc. 1st conference on Modelling Autonomic Communi-
encouraged and thus table III can become a good guide for cation Environment (MACE), Ireland, 2006.
all kinds of researchers (coming from both academy and [19] R. Schmohl and U. Baumgarten, ”Context-aware Computing: a Survey
industry) to elaborate on research aspects they are particularly Preparing a Generalized Approach”, in Proc. International Multi-
Conference of Engineers and Computer Scientists, Hong Kong, 2008.
interested in. Proposed taxonomy schemes can also be used as [20] K. Henricksen, J. Indulska, T. McFadden and S. Balasubramaniam,
a complementary tool for the better identification of the future ”Middleware for Distributed Context-Aware Systems”, Springer Verlag
CAMoWiN research challenges concerning both networking Berlin Heidelberg, LNCS 3760, pp. 846-863, 2005.
[21] R. Winter, J. H. Schiller, N. Nikaein and C. Bonnet, ”CrossTalk: Cross-
and computing oriented ICT engineering professionals. Fi- Layer Decision Support based on Global Knowledge”, IEEE Commun.
nally, multi- and inter-disciplinary research issues that have Mag. , vol. 44, no. 1, pp. 93-99, 2006.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
22 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, ACCEPTED FOR PUBLICATION
[22] L. Sarakis, G. Kormentzas and F. M. Guirao, ”Seamless Service [44] T. Gu, H. K. Pung and A. Joshi, ”A Service-Oriented Middleware for
Provision for Multi Heterogeneous Access”, IEEE Wireless Communi- Building Context-Aware Services” Elsevier Journal of Network and
cations, vol. 16, no. 5, pp. 32-40, 2009. Computer Applications (JNCA), vol. 28, no. 1, pp. 1-18, 2005.
[23] B. Jennings, S. V. Meer, S. Balasubramaniam, D. Botvich, M. [45] W. Qin, Y. Suo and Y. Shi, ”CAMPS: A Middleware for Providing
O’Foghlu, W. Donnelly and J. Strassner, ”Towards Autonomic Man- Context-Aware Services for Smart Space” in Proc. Advances in Grid
agement of Communications Networks”, IEEE Commun. Mag. , vol. and Pervasive Computing, Springer Verlag Berlin Heidelberg, LNCS
45, no. 10, pp. 112-121, 2007. 3947, pp. 644-653, 2006.
[24] J. Belschner et al., ”Optimisation of Radio Access Network Opera- [46] H. V. Kranenburg, M. S. Bargh, S. Iacob and A. Peddemors,, ”A
tion Introducing Self-x Functions: Use Cases, Algorithms, Expected Context Management Framework for Supporting Context-Aware Dis-
Efficiency Gains”, IEEE 69th Vehicular Technology Conference, VTC tributed Applications”, IEEE Commun. Mag. , vol. 44, no. 8, pp. 67-74,
Spring 2009. 2006.
[25] P. Bellavista, A. Corradi and C. Giannelli, ”A Unifying Perspective [47] P. Fahy and S. Clarke, ”CASS: Middleware for Mobile, Context-
on Context-Aware Evaluation and Management of Heterogeneous Aware Applications”, in Proc. Workshop on Context Awareness at 2nd
Wireless Connectivity”, IEEE Commun. Surveys Tutorials, vol. 13, no. International Conference on Mobile Systems, Applications and Services
3, pp. 337-357, 2011. (MobiSys ’04), Boston, 2004.
[26] K. Wrona and L. Gomez, ”Context-Aware Security and Secure [48] A. Chan and S. N. Chuang, ”MobiPADS: A Reflective Middleware for
Context-Awareness in Ubiquitous Computing Environments”, in Proc. Context-Aware Mobile Computing”, IEEE Trans. Software Engineer-
21st Conference of Polish Information Processing Society (PIPS), pp. ing, vol. 29, no. 12, pp. 1072 - 1085, 2003.
255-265, 2005. [49] The CONTEXT project, http://context.upc.es/index.htm
[27] X. Jiang and J. A. Landay, ”Modelling Privacy Control in Context- [50] J. Strassner, ”Policy-Based Network Management”, Morgan Kaufman
Aware Systems”, IEEE Pervasive Computing, vol. 1, no. 3, pp. 59-63, Publishers, ISBN 1-55860-859-1, September 2003.
2002. [51] U. Mahmud, N. Iltaf, A. Rehman and F. Kamran, ”Context-Aware
[28] J. Hong, E. H. Suh, J. Kim and S. Kim, ”Context-Aware System for Paradigm for a Pervasive Computing Environment (CAPP)”, in Proc.
Proactive Personalized Service based on Context History”, Elsevier IADIS International Conference on WWW/Internet, Villa Real, Portu-
Expert Systems with Applications Journal, vol. 36, no. 4, pp. 7448- gal, 2007, pp. 337-346.
7457, 2009. [52] R. Kodikara, C. Ahlund and A. Zaslavsky, ”Towards Context Aware
[29] H. J. La and S. D. Kim, ”A Conceptual Framework for Provisioning Adaptation in Wireless Networks”, in Proc. 2nd International Confer-
Context-Aware Mobile Cloud Services”, in Proc. IEEE 3rd Interna- ence on Mobile Ubiquitous Computing, Systems, Services and Tech-
tional Conference on Cloud Computing, 2010, pp. 466-473. nologies (UBICOMM), 2008, pp. 245-250.
[30] G. Chen and D. Kotz, ”A Survey of Context-Aware Mobile Computing [53] C. Casetti, C. F. Chiasserini, R. Fracchia and M. Meo, ”Autonomic
Research”, Dartmouth Computer Science Technical Report TR2000- Interface Selection for Mobile Wireless Users”, IEEE Trans. Veh.
381, Hanover, 2000. Technol., vol. 57, no. 6, pp. 3666-3678, 2008.
[31] E. Christopoulou, C. Goumopoulos and A. Kameas, ”An Ontology- [54] N. Saxena, A. Roy and J. Shin, ”CARP: Context-Aware Resource
Based Context Management and Reasoning Process for UbiComp Ap- Provisioning for Multimedia over 4G Wireless Networks”, in Proc.
plications”, in Proc. Joint sOc-EUSAI’2005 Conference, ACM Press, ICCS, Springer Verlag Berlin Heidelberg, LNCS 4487, pp. 652-659,
Grenoble, France, 2005, pp. 265-270. 2007.
[32] G. D. Abowd, C. G. Atkeson, J. Hong, S. Long R. Kooper and [55] A. Hasswa, N. Nasser and H. Hassanein, ”A Seamless Context-Aware
M. Pinkerton, ”Cyberguide: A Mobile Context-Aware Tour Guide”, Architecture for Fourth Generation Wireless Networks”, Springer Wire-
Baltzer/ACM Wireless Networks, vol. 3, no. 5, pp. 421-433, 1997. less Personal Communications, Special Issue on Seamless Handover
[33] N. Davies, S. Wade, A. Friday and S. B. Gordon, ”Limbo: A Tuple in next Generation Wireless/Mobile Networks, vol. 43, no. 3, pp. 1035-
Space Based Platform for Adaptive Mobile Applications”, in Proc. 1049, 2007.
International Conference on Open Distributed Processing/Distributed [56] A. De La Oliva, A. Banchs, I. Soto, T. Melia and A. Vidal, ”An
Platforms (ICODP/ICDP), Toronto, Canada, 1997, pp. 291-302. Overview of IEEE 802.21: Media-Independent Handover Services”,
[34] A. Schmidt, K. A. Aidoo, A. Takaluoma, U. Tuomela, K. V. Laerhoven IEEE Wireless Commun., vol. 15, no. 4, pp. 96-103, 2008.
and W. V. Velde, ”Advanced interaction in context”, in Proc. 1st [57] S. Buljore, H. Harada, S. Filin, P. Houze, K. Tsagkaris, O. Holland,
International Symposium on Handheld and Ubiquitous Computing K. Nolte, T. Farnham and V. Ivanov, ”Architecture and Enablers for
(HUC), Karlsruhe, Germany, 1999, pp. 89-101. Optimized Radio Resource Usage in Heterogeneous Wireless Access
[35] A. K. Dey, ”Providing Architectural Support for Building Context- Networks: The IEEE 1900.4 Working Group”, IEEE Commun. Mag. ,
Aware Applications”, PhD Thesis, College of Computing, Georgia vol. 47, no. 1, pp. 122-129, 2009.
Institute of Technology, December 2000. [58] S. Fernandes and A. Karmouch, ”Vertical Mobility Management Ar-
[36] A. K. Dey, G. D. Abowd and D. Salber, ”A Conceptual Framework chitectures in Wireless Networks: A Comprehensive Survey and Future
and a Toolkit for Supporting the Rapid Prototyping of Context-Aware Directions”, IEEE Commun. Surveys Tutorials, vol. 14, no. 1, pp. 45-
Applications”, Human-Computer Interaction, vol. 16, no. 2, pp. 97- 63, 2012.
166, 2001. [59] G. Lampropoulos, H. Marques, L. Sarakis, J. Ribeiro, C. Skianis and
[37] H. W. Gellersen, A. Schmidt and M. Beigl, ”Multi-Sensor Context- G. Kormentzas,”Enhanced Media Independent Handover Procedure for
Awareness in Mobile Devices and Smart Artifacts”, Mobile Networks Next Generation Networks”, in Proc. Future Network and Mobile
and Applications (MONET), vol. 7, no. 5, pp. 341-351, 2002. Summit 2010 Conference, Florence, Italy, 2010, pp. 1-8.
[38] P. Gray and D. Salber, ”Modelling and Using Sensed Context In- [60] Y. Wang, P. Zhang, Y. Zhou, J. Yuan, F. Liu and G. Li, ”Handover Man-
formation in the Design of Interactive Applications”, in Proc. 8th agement in Enhanced MIH Framework for Heterogeneous Wireless
IFIP International Conference on Engineering for Human-Computer Networks Environment”, Springer Wireless Personal Communications,
Interaction, LNCS 2254, 2001, pp. 317-336. vol. 52, no. 3, pp. 615-636, 2010.
[39] G. Chen and D. Kotz, ”Solar: A Pervasive-Computing Infrastructure for [61] A. Galani, K. Tsagkaris, N. Koutsouris and P. Demestichas, ”Design
Context-Aware Mobile Applications”, Technical Report TR2002-421, and Assessment of Functional Architecture for Optimized Spectrum
Dartmouth College, 2002. and Radio Resource Management in Heterogeneous Wireless Net-
[40] C. Anagnostopoulos, A. Tsounis and S. Hadjiefthymiades, ”Context works”, Wiley International Journal of Network Management, vol. 20,
Awareness in Mobile Computing Environments: a Survey”, Springer no. 4, pp. 219-241, 2010.
Wireless Personal Communications, vol. 42, no. 3, pp. 445-464, 2007. [62] A. I. Wang and Q. K. Ahmad, ”CAMF - Context-Aware Machine
[41] K. E. Kjr, ”A Survey of Context-Aware Middleware”, in Proc. 25th learning Framework for Android”, in Proc. International Conference
conference on International Multi-Conference: Software Engineering, on Software Engineering and Applications (SEA), Marina Del Rey, CA,
Innsbruck, Austria, 2007, pp. 148-155. USA, 2010.
[42] P. Debaty, P. Goddi and A. Vorban, ”Integrating the Physical World [63] L. Sliman, F. Biennier and Y. Badr, ”A Security Policy Framework for
with the Web to Enable Context-Enhanced Services”, Springer Mobile Context-Aware and User Preferences in E-Services”, Elsevier Journal
Networks and Applications, vol. 10, no. 4, pp. 385-394, 2005. of Systems Architecture, Special Issue on Secure SOA, vol. 55, no. 4,
[43] H. Chen, T. Finin and A. Joshi, ”An Ontology for Context-Aware pp. 275-288, 2009.
Pervasive Computing Environments” Knowledge Engineering Review, [64] S. Reddy, J. Burke, D. Estrin, M. Housen and M. Srivastava, ”Using
Special Issue on Ontologies for Distributed Systems, vol. 18, no. 3, pp. Mobile Phones to Determine Transportation Modes”, ACM Trans.
197-207, 2004. Sensor Networks, vol. 6, no. 2, pp. 1-27, 2010.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 23
[65] Kyungmin Cho, I. Hwang, S. Kang, B. Kim, L. Jinwon, L. Sangjeong, [87] D. Balakrishnan, M. E. Barachi, A. Karmouch and R. Glitho, ”Chal-
P. Souneil, S. Junehwa and R. Yunseok, ”HiCon: A Hierarchical lenges in Modeling and Disseminating Context Information in Ambient
Context Monitoring and Composition Framework for Next-Generation Networks”, Springer Verlag Berlin Heidelberg, LNCS 3744, pp. 32-42,
Context-Aware Services”, IEEE Network, vol. 22, no. 4, pp. 34-42, 2005.
2008. [88] 3GPP, TS 25.346 V9.1.0. ”Introduction of the multimedia broad-
[66] N. Samaan and A. Karmouch, ”Towards Autonomic Network Manage- cast/multicast service (MBMS) in the radio access network (RAN),
ment: an Analysis of Current and Future Research Directions”, IEEE stage 2”, Release 9, 2010.
Commun. Surveys Tutorials, vol. 11, no. 3, pp. 22-36, 2009. [89] ETSI EN 302 304 V1.1.1 (2004-11), Digital Video Broadcasting
[67] W. Stallings, ”SNMP and SNMPv2: The Infrastructure for Network (DVB); Transmission system for handheld terminals (DVB-H), 2004.
Management”, IEEE Commun. Mag. , vol. 36, no. 3, pp. 37-43, 1998. [90] O. Riva, T. Nadeem, C. Borcea and L. Iftode, ”Context-Aware Migra-
[68] S. J. Yoo and N. Golmie, ”Policy-based scanning with QoS support for tory Services in Ad Hoc Networks”, IEEE Trans. Mobile Computing,
seamless handovers in wireless networks”, Wireless Communications vol. 6, no. 12, pp. 1313-1328, 2007.
and Mobile Computing, vol. 10, no. 3, pp. 405-425, 2010. [91] Y. Vanrompay, A. H. Yasar, D. Preuveneers and Y. Berbers, ”Context-
[69] E. Linehan, S. L. Tsang and S. Clarke, ”Supporting Context- aware optimized information dissemination in large scale vehicular
Awareness: A Taxonomic Review”, TCD-CS-2008-37, October 2008. networks”, in Proc. 3rd ACM International Conference on Pervasive
[70] S. Kang, L. Jinwon, J. Hyukjae, L. Youngki, P. Souneil and S. Junehwa, Technologies Related to Assistive Environments (PETRA), Samos,
”A Scalable and Energy-Efficient Context Monitoring Framework for Greece, 2010.
Mobile Personal Sensor Networks”, IEEE Trans. Mobile Computing, [92] M. Barachi, A. Kadiwal, R. Glitho, F. Khendek and R. Dssouli, ”The
vol. 9, no. 5, pp. 686-702, 2010. design and implementation of architectural components for the integra-
[71] P. Nurmi, M. Martin and J. A. Flanagan, ”Enabling Proactiveness tion of the IP multimedia subsystem and wireless sensor networks”,
Through Context Prediction”, in Proc. Workshop on Context Awareness IEEE Commun. Mag. , vol. 48, no. 4, pp. 42-50, 2010.
for Proactive Systems (CAPS), Helsinki University Press, 2005, pp. [93] J. Antoniou, F. C. Pinto, J. Simoes and A. Pitsillides, ”Supporting
159-168. Context-Aware Multiparty Sessions in Heterogeneous Mobile Net-
[72] T. Anagnostopoulos, C. Anagnostopoulos, S. Hadjiefthymiades, M. works” Elsevier Mobile Networks and Applications, vol. 15, no. 6,
Kyriakakos and A. Kalousis, ”Predicting the Location of Mobile pp. 831-844, 2010.
Users: A Machine Learning Approach”, in Proc. ACM International [94] Xiaoming Hu et al., ”A Hybrid Peer-to-Peer Solution for Context
Conference on Pervasive Services (ICPS), Imperial College, London, Distribution in Mobile and Ubiquitous Environments”, Springer In-
2009. formation Systems Development, Paphos, Cyprus, pp. 501-510, 2009.
[73] R. Mayrhofer, H. Radi and A. Ferscha, ”Recognizing and Predicting [95] K. Stanoevska-Slabeva, T. Wozniak, I. Hoffend, C. Mannweiler, H. D.
Context by Learning from User Behavior”, Radiomatics: Journal of Schotten, ”The Emerging Ecosystem for Context Information and the
Communication Engineering, Special Issue on Advances in Mobile Role of Telecom Operators”, in Proc. 14th International Conference
Multimedia, vol. 1, no. 1, pp. 30-42, 2004. on Intelligence in Next Generation Networks (ICIN), 2010.
[74] S. Sigg, S. Haseloff, and K. David, ”An Alignment Approach for [96] A. Galani, K. Tsagkaris and P. Demestichas, ”Information Flow for
Context Prediction Tasks in UbiComp Environments,” IEEE Pervasive Optimized Management of Spectrum and Radio Resources in Cognitive
Computing, vol. 9, no. 4, pp. 90-97, 2010. B3G Wireless Networks”, Springer Network and Systems Management,
[75] Y. Wang, M. Martonosi and L. S. Peh, ”Predicting Link Quality using vol. 18, no. 2, pp. 125-149, 2010.
Supervised Learning in Wireless Sensor Networks”, Mobile Computing [97] IEEE P1900.4/D1.5 (2008), ”Draft Standard for Architectural Building
and Communications Review (MC2R), July, 2007. Blocks Enabling Network-device Distributed Decision Making for
[76] A. Krause, A. Smailagic and D. P. Siewiorek, ”Context-Aware Mobile Optimized Radio Resource Usage in Heterogeneous Wireless Access
Computing: Learning Context- Dependent Personal Preferences from Networks”, IEEE.
a Wearable Sensor Array”, IEEE Trans. Mobile Computing, vol. 5, no. [98] IEEE Std 802.21-2008, (2009), ”IEEE Standard for Local and
2, pp. 113- 127, 2006. Metropolitan Area Networks-Part 21: Media Independent Handover
[77] Xiaojin Zhu, ”Semi-Supervised Learning”, In Claude Sammut and Services”, IEEE.
Geoffrey Webb, editors, Encyclopedia of Machine Learning, Springer, [99] S. L. Kiani, M. Knappmeyery, N. Baker, and B. Moltchanov, ”A
to appear. Federated Broker Architecture for Large Scale Context Dissemination”,
[78] C. Bettini, O. Brdiczka, K. Henricksen, J. Indulska, D. Nicklas, A. in Proc. 10th IEEE International Conference on Computer and Infor-
Ranganathan and D. Riboni, ”A Survey of Context Modelling and mation Technology (CIT 2010), Bradford, UK, 2010, pp. 2964-2969.
Reasoning Techniques”, Elsevier Pervasive and Mobile Computing, [100] A. Klein, C. Mannweiler, J. Schneider and H. D. Schotten, ”Access
vol. 6, no. 2, pp. 161-180, 2010. Schemes for Mobile Cloud Computing,” in Proc. 11th International
[79] B. Moltchanov, C. Mannweiler and J. Simoes, ”Context-Awareness Conference on Mobile Data Management (MDM), May 2010, pp. 387-
Enabling New Business Models in Smart Spaces”, Springer Verlag 392.
Berlin Heidelberg, LNCS 6294, pp. 13-25, 2010. [101] J. Antoniou, C. Christophorou, J. Simoes and A. Pitsillides, ”Adaptive
[80] Y. M. Park, A. Moon, Y. Choi, S. K. Kim and S. Kim, ”An Efficient Network-Aided Session Support in Context-Aware Converged Mobile
Context Model for Fast Responsiveness of Context-Aware Services in Networks”, International Journal of Autonomous and Adaptive Com-
Mobile Networks”, in Proc. 7th IEEE Consumer Communications and munication Systems, March 2010.
Networking Conference (CCNC), Las Vegas, Nevada, USA, January [102] J. Simes and T. Magedanz, ”Contextualized User-Centric Multimedia
2010. Delivery System for Next Generation Networks”, Springer Telecom-
[81] A. Agostini, C. Bettini and D. Riboni, ”Hybrid Reasoning in the munication Systems, vol. 48, no. 3-4, pp. 301-316, 2011.
CARE Middleware for Context-Awareness”, International Journal of [103] A. U. H. Yasar, Y. Vanrompay, D. Preuveneers and Y. Berbers, ”Opti-
Web Engineering and Technology, vol. 5, no. 1, pp. 3-23, 2009. mizing Information Dissemination in Large Scale Mobile Peer-to-Peer
[82] K. Henricksen, S. Livingstone and J. Indulska, ”Towards a Hybrid Networks Using Context-based Grouping”, in Proc. 13th International
Approach to Context Modeling, Reasoning and Interoperation”, in: J. Conference on Intelligent Transportation Systems, 2010, pp. 1065-
Indulska, D.D. Roure (Eds.), Proc. First International Workshop on 1071.
Advanced Context Modeling, Reasoning and Management, University [104] P. Makris, G. Lampropoulos, D. Skoutas and C. Skianis ”New Di-
of Southampton, Nottingham, England, 2004. rections and Challenges for MIH Operation Towards Real Market
[83] B. A. Truong, Y. K. Lee and S. Y. Lee, ”Modeling and Reasoning about Applicability”, in Proc. 1st International Workshop on Mobility in
Uncertainty in Context-Aware Systems”, in Proc. IEEE International Future Internet (MiFI ’10), Crete Greece, 2010.
Conference on e-Business Engineering (ICEBE), 2005, pp. 102-109. [105] N. Dimitriou, L. Sarakis, D. Loukatos, G. Kormentzas and C. Skianis,
[84] H. Chen, F. Perich, T. Finin and A. Joshi, ”SOUPA: Standard Ontology ”Vertical Handover Framework for Future Collaborative Wireless Net-
for Ubiquitous and Pervasive Applications”, in Proc. 1st Annual Inter- works”, Wiley International Journal of Network Management, vol. 21,
national Conference on Mobile and Ubiquitous Systems, MobiQuitous pp. 548-564, 2011.
2004, IEEE Computer Society, 2004. [106] P. Demestichas, ”Introducing Cognitive Systems in the Wireless
[85] B. Siljee, I. Bosloper, and J. Nijhuis, ”A Classification Framework for B3G World: Motivations and Basic Engineering Challenges”, Elsevier
the Storage and Retrieval of Context”, in Proc. Workshop on Modelling Telematics and Informatics, vol. 27, no. 3, pp. 256-268, 2010.
and Retrieval of Context (MRC), Ulm, Germany, 2004. [107] A. Saatsakis, K. Tsagkaris and P. Demestichas, ”Exploiting Context,
[86] J. Coutaz, J. L. Crowley, S. Dobson and D. Garlan, ”Context is Key”, Profiles and Policies in Dynamic Sub-carrier Assignment Algorithms
ACM Communications Magazine, vol. 48, no. 3, pp. 49-53, 2005. for Efficient Radio Resource Management in OFDMA Networks”,
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
24 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, ACCEPTED FOR PUBLICATION
Springer Annals of Telecommunications, vol. 65, no. 7, pp. 359-374, [131] T. W. Wlodarczyk, and C. Rong, ”On the Sustainability Impacts
2010. of Cloud-Enabled Cyber Physical Space”, in Proc. 2nd IEEE Inter-
[108] K. Nolte et al., ”The E3 Architecture: Enabling Future Cellular national Conference on Cloud Computing Technology and Science
Networks with Cognitive and Self-x Capabilities”, Wiley International (CloudCom), 2010, pp. 597-602.
Journal of Network Management, vol. 21, no. 5, pp. 360-383, 2011. [132] L. Hilty and M. Hercheui, ”ICT and Sustainable Development,” What
[109] H. Derbel, N. Agoulmine and M. Salan, ”ANEMA: Autonomic Net- Kind of Information Society? Governance, Virtuality, Surveillance,
work Management Architecture to Support Self-Configuration and Sustainability, Resilience, Springer Boston, 2010, pp. 227-235.
Self-Optimization in IP Networks, Elsevier Computer Networks, vol. [133] ”Google CEO Schmidt: ”People Aren’t Ready
53, no. 3, pp. 418-430, 2009. for the Technology Revolution”” Available:
[110] H. J. Wang, R. H. Katz and J. Giese, ”Policy-Enabled Handoffs Across http://www.readwriteweb.com/archives/google ceo schmidt people are
Heterogeneous Wireless Networks”, in Proc. 2nd IEEE Workshops on nt ready for the tech.php
Mobile Computer Systems and Applications (WMCSA), 1999. [134] K. C. Laudon, ”Ethical Concepts and Information Technology,” Com-
[111] E . C. Lupu and M. S. Sloman, ”Conflicts in Policy-Based Distributed munications of the ACM, vol. 38, no. 12, pp. 33-40, 1995.
Systems Management”, IEEE Trans. Software Engineering - Special [135] J. Saltzer and M. Schroeder, ”The Protection of Information in Com-
Issue on Inconsistency Management, vol. 25, no. 6, pp. 852-869, 1999. puter Systems”, IEEE Computer Society Press, vol. 63, no. 9, pp. 1278-
[112] P. Makris and C. Skianis, ”Multi-Scenario Based Call Admission 1308, 1975.
Control for Coexisting Heterogeneous Wireless Technologies”, in Proc. [136] ”Trust in the Information Society”, A Report of the Advisory Board
IEEE Global Telecommunications Conference (GLOBECOM), New for Research & Innovation on Security, Privacy and Trustworthiness in
Orleans, USA, 2008. the Information Society (RISEPTIS), 2010.
[113] A. Rahmati and L. Zhong, ”Context-Based Network Estimation for [137] Mark Stamp, ”Information Security Principles and Practice”, John
Energy-Efficient Ubiquitous Wireless Connectivity”, IEEE Trans. Mo- Wiley & Sons Inc., ISBN: 978-0471738480, 2006.
bile Computing, vol. 10, no. 1, pp. 54-66, 2011. [138] L. Gomez, A. Laube and A. Sorniotti, ”Access Control in Mobile
[114] A. Izquierdo and N. T. Golmie, ”Improving Security Information and Ubiquitous Environments” in D. Stojanovic book ”Context-Aware
Gathering with IEEE 802.21 to Optimize Handover Performance”, Mobile and Ubiquitous Computing for Enhanced Usability: Adaptive
in Proc. 12th international conference on Modelling, Analysis and Technologies and Applications”, Information Science Reference, pp.
Simulation of Wireless and Mobile Systems (MSWiM), 2009, pp. 96- 278-294, 2009.
105. [139] J. Cho, A. Swami, and I. Chen, ”A Survey on Trust Management for
[115] J. Zander, ”Radio Resource Management in Future Wireless Networks: Mobile Ad Hoc Networks,” IEEE Commun. Surveys Tutorials, vol. 13,
Requirements and Limitations”, IEEE Commun. Mag., vol. 35, no. 8, no. 4, pp. 562-583, 2011.
pp. 30-36, 1997. [140] Emin Islam Tatl?, ”Security in Context-aware Mobile Business Appli-
[116] D. Niyato and E. Hossain, ”Call Admission Control for QoS Provision- cations”, PhD Thesis, Mannheim University, 2009.
ing in 4G Wireless Networks: Issues and Approaches”, IEEE Network, [141] P. Abi-Char, A. M’hamed, B. EL-Hassan and M. Mokhtari, ”Con-
vol. 19, no. 5, pp. 5-11, 2005. trolling Trust and Privacy in Context-Aware Environments, State of
[117] Y. Wang, P. Zhang, Y. Zhou, J. Yuan, F. Liu and G. Li, ”Handover Man- Art and Future Directions”, in Proc. IGI Publishing under Book Title:
agement in Enhanced MIH Framework for Heterogeneous Wireless Trust Modeling and management in Digital Environments: From Social
Networks Environment”, Springer Wireless Personal Communications, Concept to System Development, Book Edited by Nokia Research
vol. 52, no. 3, pp. 615-636, 2010. Center, pp. 352-377, Finland, 2010.
[118] E. Patouni, N. Alonistioti, and L. Merakos, ”Modeling and Perfor- [142] M. Tschersich, C. Kahl, S. Heim, S. Crane, K. Bottcher, I. Krontiris
mance Evaluation of Reconfiguration Decision Making in Heteroge- and K. Rannenberg, ”Towards Privacy-enhanced Mobile Communities
neous Radio Network Environments” IEEE Trans. Veh. Technol., vol. - Architecture, Concepts and User Trials”, Elsevier Journal of Systems
59, no. 4, pp. 1887-1900, 2010. and Software, vol. 84, no. 11, pp. 1947-1960, 2011.
[119] D. N. Skoutas and A. N. Rouskas, ”A Scheduling Algorithm with [143] E. Aivaloglou, S. Gritzalis and C. Skianis, ”Requirements and Chal-
Dynamic Priority Assignment for WCDMA Systems”, IEEE Trans. lenges in the Design of Privacy-aware Sensor Networks”, in Proc. 49th
Mobile Computing, vol. 8, no. 1, pp. 126-138, 2009. IEEE GLOBECOM Conference, San Francisco, USA, 2006.
[120] J. O. Kephart and D. M. Chess, ”The Vision of Autonomic Computing,” [144] A. Zugenmaier and A. Hohl, ”Anonymity for Users of Ubiquitous
Computer, vol. 36, no. 1, pp. 41-50, 2003. Computing”, in Proc. 2nd Workshop on Security in Ubiquitous Com-
[121] A. Saatsakis and P. Demestichas, ”Context Matching for Realizing puting (Ubicomp), USA, 2003.
Cognitive Wireless Networks Segments”, Springer Wireless Personal [145] L. Wijngaert and H. Bouwman, ”Would you share? Predicting the
Communications, vol. 55, no. 3, pp. 407-440, 2010. potential use of a new technology”, Telematics and Informatics, vol.
[122] D. Riecken, ”Personalized Views of Personalization”, Communications 26, no. 1, pp. 85-102, 2009.
of the ACM, vol. 43, no. 8, pp. 26-28, 2000. [146] V. Cahill et al., ”Using Trust for Secure Collaboration in Uncertain
[123] D. N. Skoutas, P. Makris and C. Skianis, ”Optimized Admission Environments”, IEEE Pervasive Computing, Special Issue on Dealing
Control Scheme for Coexisting Femtocell, Wireless and Wireline with Uncertainty, vol. 2, no. 3, pp. 52-61, 2003.
Networks”, Springer Telecommunication Systems, Special Issue on [147] M. S. Lund, B. Solhaug and K. Stlen, ”Evolution in Relation to Risk
Mobility Management in Future Internet, in press. and Trust Management”, IEEE Computer Society, vol. 43, no. 5, pp.
[124] M. D. Reuver and T. Haaker, ”Designing Viable Business Models for 49-55, 2010.
Context-Aware Mobile Services”, Elsevier Telematics and Informatics, [148] B. G. Chun and P. Maniatis, ”Dynamically Partitioning Applications
vol. 26, no. 3, pp. 240-248, 2009. Between Weak Devices and Clouds”, in Proc. 1st ACM Workshop on
[125] M. Satyanarayanan, P. Bahl, R. Caceres and N. Davies, ”The Case for Mobile Cloud Computing and Services (MCS), San Francisco, USA,
VM-based Cloudlets in Mobile Computing”, IEEE Pervasive Comput- 2010.
ing, vol. 8, no. 4, pp. 14-23, 2009. [149] P. Simoens, F. De Turck, B. Dhoedt and P. Demeester, ”Remote Display
[126] L. Badia, M. Levorato, F. Librino and M. Zorzi, ”Cooperation Tech- Solutions for Mobile Cloud Computing”, IEEE Computer, vol. 44, no.
niques for Wireless Systems from a Networking Perspective,” IEEE 8, pp. 46-53, 2011.
Wireless Commun., vol. 17, no. 2, pp. 89-96, 2010. [150] B. Ahlgren, P. A. Aranda, P. Chemouil, S. Oueslati, L. M. Correia,
[127] A. Jameson, ”Modeling both the context and the user”, Personal and H. Karl, M. Sllner and A. Welin, ”Content, Connectivity, and Cloud:
Ubiquitous Computing, vol. 5, no. 1, pp. 29-33, 2001. Ingredients for the Network of the Future”, IEEE Commun. Mag., vol.
[128] G. Adomavicius and A. Tuzhilin, ”Personalization Technologies: A 49, no. 7, pp. 62-70, 2011.
Process-Oriented Perspective”, Communications of the ACM, vol. 48, [151] K. Kumar and Y. Lu, ”Cloud Computing for Mobile Users: Can
no. 10, pp. 83-90, 2005. Offloading Computation Save Energy?”, IEEE Computer, vol. 43, no.
[129] G. Adomavicius and A. Tuzhilin, ”Toward the Next Generation of 4, pp. 51-56, 2010.
Recommender Systems: A Survey of the State-of-the-art and Possible [152] Y. Lu, S. Li and H. Shen, ”Virtualized Screen: A Third Element for
Extensions”, IEEE Trans. Knowl. Data Eng., vol. 17, no. 6, pp. 734- Cloud-Mobile Convergence”, IEEE Multimedia, vol. 18, no. 2, pp. 4-
749, 2005. 11, 2011.
[130] R. L. Aguiar, A. Sarma, D. Bijwaard, L. Marchetti, and P. Pacyna, [153] D. Huang, Z. Zhou, L. Xu, T. Xing and Y. Zhong, ”Secure data
”Pervasiveness in a Competitive Multi-Operator Environment: The processing framework for mobile cloud computing”, in Proc. IEEE
Daidalos Project”, IEEE Commun. Mag. , vol. 45, no. 10, pp. 22-26, Conference on Computer Communications Workshops (INFOCOM),
2007. 2011, pp. 614-618.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
MAKRIS et al.: SURVEY ON CONTEXT-AWARE MOBILE AND WIRELESS NETWORKING: NETWORKING/COMPUTING ENVIRONMENT INTEGRATION 25
[154] H. Liang, D. Huang, L. X. Cai, X. Shen and D. Peng, ”Resource alloca- Dimitrios N. Skoutas is currently a member of
tion for security services in mobile cloud computing”, in Proc. IEEE Computer and Communication Systems Laboratory
Conference on Computer Communications Workshops (INFOCOM), (CCSL) a research and educational unit in the De-
2011, pp. 191-195. partment of Information and Communication Sys-
[155] F. Toutain, A. Bouabdallah, R. Zemek, and C. Daloz, ”Interpersonal tems Engineering (ICSE) of the University of the
Context-Aware Communication Services”, IEEE Commun. Mag., vol. Aegean, Greece. He holds a PhD degree in Commu-
49, no. 1, pp. 68-74, 2011. nication Networks and a Dipl.-Eng (5 years degree)
[156] M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, K. in Electrical and Computer Engineering with major
Young-Doo, K. Eungsun and C. Yoon-Chae, ”An Overview of Radio in Telecommunications.
Resource Management in Relay-Enhanced OFDMA-Based Networks”, His current research activities include context
IEEE Commun. Surveys Tutorials, vol. 12, no. 3, pp. 422-438, 2010. aware Next Generation Networks and Quality of
[157] B. Wang and K. J. R. Liu, ”Advances in Cognitive Radio Networks: A Service provisioning in heterogeneous networks environment. He has also
Survey”, IEEE J. Sel. Topics Signal Process., vol. 5, no. 1, pp. 5-23, been keenly working on the area of Resource Management and Quality of
2011. Service Provisioning in Mobile and Wireless Broadband Networks where
[158] J. Brnsted, K. M. Hansen and M. Ingstrup, ”Service Composition Issues he has proposed several algorithmic and architectural optimizations. He has
in Pervasive Computing”, IEEE Pervasive Computing, vol. 9, no. 1, pp. several publications in the above fields in international scientific journals
62-70, 2010. and conference proceedings. Dr. Skoutas has also been involved in various
[159] E. Trouva, E. Grasa, J. Day, I. Matta, L. Chitkushev, S. Bunch, European and National funded R&D projects (e.g. FP7-ICT-PASSIVE, FP6-
M. P. de Leon, P. Phelan and X. Hesselbach-Serra, ”Transport over IST-UNITE, ESF-NR ARCHIMEDES I).
Heterogeneous Networks Using the RINA Architecture”, in Proc. 9th He is at the editorial board of ACM/Springer Wireless Networks (WINET),
International Conference on Wired/Wireless Internet Communications he acts within Technical Program and Organizing Committees for conferences
(WWIC), 2011, pp. 297-308. and he is an active reviewer for several scientific journals and conferences.
[160] P. Marsch, B. Raaf, A. Szufarska, P. Mogensen, H. Guan, M. Farber, S. He is a member of pronounced professional societies (IEEE Communications
Redana, K. Pedersen and T. Kolding, ”Future Mobile Communication Society, Greek Computer Society and Technical Chamber of Greece).
Networks: Challenges in the Design and Operation”, IEEE Vehicular
Technology Magazine, vol. 7, no. 1, pp. 16-23, 2012.
[161] X. Wang, A. Vasilakos, M. Chen, Y. Liu and T. T. Kwon, ”A
Survey of Green Mobile Networks: Opportunities and Challenges”,
ACM/Springer Mobile Networks and Applications (MONET), vol. 17,
no. 1, pp. 4-20, 2012. Charalabos Skianis (Senior Member IEEE) is cur-
[162] S. Pearson, ”Toward Accountability in the Cloud”, IEEE Internet rently director of computer & Communication Sys-
Computing, vol. 15, no. 4, , pp. 64-69, 2011. tems Lab. and an assistant professor in the Depart-
[163] W. Jansen and T. Grance, ”Guidelines on Security and Privacy in ment of Information and Communication Systems
Public Cloud Computing”, NIST: National Institute of Standards and at the University of the Aegean, Samos, Greece.
Technology, Technical Report 800-144, 2011. He holds a Ph.D. degree in computer science, Uni-
[164] Z. Ahmad, K. E. Mayes, S. Dong and K. Markantonakis, ”Consider- versity of Bradford, United Kingdom, and a B.Sc.
ations for mobile authentication in the Cloud”, Elsevier Information in physics, Department of Physics, University of
Security Technical Report, pp. 123-130, 2011. Patras, Greece.
His work is published in journals, conference
proceedings and as book chapters and has also
been presented in numerous conferences and workshops. He acts within
Technical Program and Organizing Committees for numerous conferences
Prodromos Makris was born in Samos Island, and workshops (e.g., IFIP Networking, symposium chair for several IEEE
Greece, in 1985. He holds a five-year Diploma Globecom and IEEE ICC conferences) and as a Guest Editor for scientific
in Information & Communication Systems Engi- journals (e.g., IEEE Networks magazine).
neering (2007), and an M.Sc. in Communication He is at the editorial board of journals (e.g., IEEE Commun. Mag. , IEEE
& Computer Networking Technologies (2009), both Surveys & Tutorials, Ad hoc Networks), a member of pronounced professional
from the University of the Aegean, Greece. societies and an active reviewer for several scientific journals. He is an active
He is currently pursuing his Ph.D. at the same member of several Technical Committees within the IEEE ComSoc (e.g.,
university and his research interests include context- chair for the CSIM technical committee, past vice chair for Information
aware mobile & wireless networking, QoS provi- Infrastructure technical committee), member of the IEEE EMEA board, the
sioning in 4G heterogeneous network technologies education board of IEEE ComSoc, and member of IEEE BTS; IEEE TVT and
and mobile cloud computing issues. Mr. Makris IEEE CS. He has acted as Technical Manager for ICT FP7 VITAL++ project
is also an IEEE Graduate Student member and member of the Technical and representative for the University on ICT FP7 HURRICANE project. He
Chamber of Greece. is currently project coordinator for ICT FP7 PASSIVE project.