SECURE WEARABLE AND IMPLANTABLE BODY SENSOR
NETWORKS IN HAZARDOUS ENVIRONMENTS
Mohamed Hamdi, Noureddine Boudriga
Communications Networks and Security Research Lab., Ariana, Tunisia
Habtamu Abie
Norwegian Computing Center, Oslo, Norway
Mieso Denko
University of Guelph, Ontario, Canada
Keywords: Smart sensor networks, Wearable and implantable sensors, Intelligent session management, Secure
communication.
Abstract: The aim of Wearable and implantable monitoring devices is to collect relevant data from the application-
related environment, and transmit this information to the outside world. Modern microelectronics create
ever increasing opportunities, but it is still true that sensors form the weakest elements in the entire chain of
data collection and processing. The difficulty of deploying smart body sensor networks is exacerbated by
the hostile environments in which they are typically installed. In this paper, we propose a novel architecture
for wearable and implantable body sensor systems that guarantees both real-time responsiveness and
security. We rely on the wavelet packet transform to develop an intelligent session management scheme
where a customizable frame structure allows multiplexing the set of sessions between the elementary
sensors and the analysis center. We introduce a lightweight identity-based encryption protocol suitable for
body smart sensor systems. We also present performance results using simulation experiments.
1 INTRODUCTION
The last decade has witnessed a rapid surge of interest
in new sensing and monitoring devices for healthcare
and the use of wearable/wireless devices for multiple
applications (Puers, 2005). Key developments in this
area include implantable in vivo monitoring,
battlefield monitoring, and human tracking; where
sensors are strategically placed at various locations on
the vest or inside the human body to form a network
(Body Sensor Network) and interact with the human
system to acquire and transmit the data to an
acquisition system. The data acquisition hardware
collects the data from various sensors and transmits
the processed data to the remote monitoring station.
The basic requirement for such systems is that the
data gathered by the body sensors should be available
for transmission in real-time in response to a query
issued by the data acquisition system. When multiple
sensors are involved in the measurement process,
real-time responsiveness becomes hard to achieve
since all sensor nodes share the same communication
channel. Hence, the balance between response delay
and scalability should be carefully addressed. In
addition, security is a matter of concern in these
networks, as the data being monitored are the health
status of the individual. The sensor nodes used to
form these networks are resource-constrained, which
makes security applications a challenging problem.
The data are also vulnerable to external attackers,
who may inject errors in the routing information,
replay old routing information, distort routing
information or send malicious information. The data
are also subject to jamming, tampering, Sybil attack,
and collision (Hamdi and Boudriga, 2008. Attacks of
this nature which have been thoroughly investigated
and neutralized as threats within the context of
traditional wireless sensor networks, still represent a
threat to body sensor networks.
85
Hamdi M., Boudriga N., Abje H. and Denko M. (2010).
SECURE WEARABLE AND IMPLANTABLE BODY SENSOR NETWORKS IN HAZARDOUS ENVIRONMENTS.
In Proceedings of the International Conference on Data Communication Networking and Optical Communication Systems, pages 85-92
DOI: 10.5220/0002988200850092
Copyright
c
SciTePress
In this paper we develop a secure network
architecture for Wearable and Implantable Smart
Sensor Networks (WISSNs). We first propose an
architecture where intermediate sensor nodes allow
the data collected by the elementary sensors to be
forwarded to the analysis center (i.e., data acquisition
center). To support this architecture, we propose a
session multiplexing scheme that permits multiple
elementary sensors to share the communication
resources of an intermediate. It relies on wavelet
theory since it is necessary to allocate a variable
number of slots to a given elementary sensor in the
multiplexed frames, the number of slots varying with
the volume of date it generates or with its residual
amount of energy. We propose, in addition, a security
protocol addressing specific issues including
authentication/anonymity, accounting, confidentiality,
and investigation. The use of elliptic curve
cryptography minimizes the power consumption of
the cryptographic primitives while the absence of user
information in the authentication protocol preserves
privacy and anonymity.
The reader will notice that four innovative issues
are addressed in this paper:
Layer-2 Multiplexing: Rather than being performed
at the physical layer, the multiplexing of the
information originating from multiple sensor nodes
is dealt with using a specific frame structure based
on the wavelet packet transform. Such multiplexing
provides more fairness.
Real-time Responsiveness: The proposed
architecture guarantees that the queries from the
data acquisition center are processed in real-time by
the smart sensor system since a set of intermediate
sensors processes the information collected by the
elementary sensors
Privacy/anonymity Provision: Our security
protocol allows the data sent by the intermediate
sensors to the analysis center to be enciphered using
dynamic public keys. This guarantees the privacy of
the collected data as well as the anonymity of the
wearer.
Low Energy Consumption: The simulations that
have been performed show that our cryptographic
protocol is characterized by a low computational
complexity, making it convenient for use with the
limited resources of the intermediate sensors
The rest of this paper is organized as follows. Section
II presents related work. Section III presents the
WISSN. A novel session multiplexing technique
based on wavelet theory is discussed in Section IV.
Section V discusses security issues. Section VI
provides validation and performance evaluation of the
proposed techniques. And Finally, Section VII
concludes the paper with suggested avenues of future
research.
2 RELATED WORK
In recent years, there has been a proliferation of smart
monitoring based on small sensing devices. A large
portion of these devices have been devised for sports
science and combating obesity. For instance, there are
sophisticated watches available today (Polar),
(Suunto) that provide real-time measurement of
heartrate and allow athletes to store the gathered data
on computers for further analysis using specific
software. Bodymedia (BodyMedia) has developed an
armband that has multiple sensors (galvanic skin
response, skin and near-body temperature, two-axis
accelerometer and heat flux) and collects
physiological data on an on-going basis for days at a
time. Once the data is uploaded to a computer,
relevant and accurate information can be extracted
about, for example, fatigue, duration of physical
activity, consumed calories, etc. However, in all cases
the physiological data is analyzed on a home PC at a
later time, and proprietary data formats prevent users
from consolidating and correlating health monitoring
data from different devices.
In the medical domain, research is being
conducted on the remote monitoring of physiological
reactions (Scannell et al., 1995), (Martin et al., 2000),
(Oliver et al., 2006). However, in existing approaches,
as a rule no automated analysis is performed by the
device, and the raw data is instead sent to a remote
computer for further analysis by humans.
Traditionally, personal medical monitoring systems,
such as Holter monitors, have been used only to
collect data for off-line processing. An exception to
this is the approach proposed in (Oliver et al., 2006)
where a cell phone is used to store, transmit (via
Bluetooth) and analyze the physiological data, and
present it to the user in an intelligible way. In (Leister
et al., 2009) a security and authentication architecture
using MPEG-21 for wireless patient monitoring
systems has been developed based on the threat
assessment of wireless patient monitoring systems. In
(Leister et al., 2008), an architecture that can handle
end-to-end management of multimedia content in
diverse wireless sensor networks have been
proposed.
Martin et al discuss in (Martin et al., 2000) the
usage of wearable computers for health monitoring
where the devices provide real-time feedback to the
patient. In particular, they describe a wearable ECG
DCNET 2010 - International Conference on Data Communication Networking
86
device, but provide no experimental results. A
wearable health-monitoring device using a Personal
Area Network (PAN) or Body Area Network (BAN)
can be integrated into a user’s clothing (Park and
Jayaraman, 2003), like Foster-Miller’s health
monitoring garment for soldiers. Along these lines,
Paradiso (Paradiso, 2003) describes preliminary work
on the WEALTHY system, a garment with embedded
ECG sensors for continuous monitoring of the heart.
Jovanov et al present in (Jovanov et al., 2005) a
wireless BAN with motion sensors for computer-
assisted physical rehabilitation and ambulatory
monitoring. In (Kemp et al., 2008), Kamp et al
develop a wearable system for manned bomb disposal
missions. Mihovska and Prasad (Mihovska and
Prasad, 2007) have developed an adaptive security
architecture for personal networks with an
asymmetric key agreement scheme on three levels by
using contextual information, such as the location of
the user and the capability of the devices. This
architecture is based on an elliptic curve
cryptosystem. It has, however, one shortcoming. It is
susceptible to impersonation via key compromise.
A global notice about these approaches shows that
traditional communication protocols are used to
transmit the collected data from the human body to an
external system (e.g., cellphone, laptop).
Unfortunately, this does not guarantee a real-time
transmission of this information since an important
variable delay can occur, especially when some
sensors transmit large units of data such as images.
Moreover, due to the use of radio communication, the
confidentiality of the transmitted data is not
intrinsically guaranteed, which may lead to privacy
violation. In several applications, including
healthcare, even the identity of the wearer should be
hidden.
3 PROPOSED WISSN
ARCHITECTURE
In this paper, we address two crucial issues regarding
wearable sensor systems:
Improving Real-time Responsiveness: This is
achieved by building special communication frame
structures based on the non-uniform multiplexing of
the data generated by different types of sensors
Combining Sensor Authentication and user
anonymity through the use of lightweight
cryptographic protocols: In order to adapt to the
severe resource limitations characterizing WISSNs,
we use an elliptic curve implementation of the
proposed security functions
In spite of its apparent simplicity, WISSNs exhibit
several complex features and therefore require
sophisticated engineering approaches in order to be
set up. In the following, we list the most relevant
factors that may shape the communication models
used in smart sensor networks.
1. Multi-functional framework: A sensor node
may be able to carry out multiple functions that can
be set on/off depending on the situation. Obviously,
the communication requirements may differ greatly
from one functionality to another according to the
data sent across the WISSN. For instance, when the
network is deployed in a mining structure, a first
category of sensor may be used to monitor the amount
of several toxic gases in the atmosphere. A second
type of sensor can serve to estimate the opacity of the
encountered obstacles. IRM sensors can be used in
such a context in order to predict, and possibly
prevent, disasters. Since the volume of data generated
by the latter category is by far greater than that
generated by the former, much more bandwidth must
be reserved to transmit image data.
2. Independent monitoring capability: Due to the
non-uniform nature of the monitored events
(irrespective of the application), some sensors may
exhaust their energy more rapidly than others. This
may result in the presence of uncovered regions
where the nodes in charge of gathering data related to
the environment are out of power. Since such a
situation significantly affects the efficiency of the
WISSN, solutions should be proposed to avoid it. One
alternative is to tune the quality of the data gathered
by a sensor node according to its residual energy
resources. This would extend considerably the
lifetime of this node at the cost of losing some refined
data, which is definitely better than totally losing the
functionalities provided by the node. As a result the
communication resources required to transmit the
data may vary from one sensor to another.
3. Exportable configuration: Configurations can
be exported from one sensor to another in order to
turn on/off several functionalities. Even though this
feature allows energy to be saved (by triggering
power-consuming time only when necessary), it
creates a significant security hole since node
imposture can be easily carried out. Hence,
authentication mechanisms should be set up to
prevent non-authorized nodes from manipulating the
WISSN. Two important issues must be taken into
consideration: First, the security algorithms must be
based on non-complex algorithms and use small
cryptographic credentials (to adapt to limited CPU
time and memory resources) and; Second For a wide
range of applications, the anonymity of the person
holding the wearable or implantable smart sensor
system should be preserved. Since this conflicts with
authentication, specific security infrastructures will
SECURE WEARABLE AND IMPLANTABLE BODY SENSOR NETWORKS IN HAZARDOUS ENVIRONMENTS
87
have to be developed
From the foregoing discussion, it transpires that
sessions on WISSNs should be managed bearing in
mind that the specific features of such networks. In
fact, a session should typically be initiated by a
central external (i.e., not wearable) node, called the
analysis center, in order to collect data from the
WISSN. Henceforth, the underlying bandwidth
management scheme should guarantee fairness for all
body sensors. Unlike traditional networks, fairness in
WISSNs should take into account the differences in
the nature of the generated data and the available
power level.
Figure 1: Wavelet packet decomposition.
Figure 1 illustrates this reasoning. In order to improve
the scalability of the communication structure, we
propose to divide the body area network into a set of
clusters. We make the assumption that within each of
these clusters, there is a central node which is in
charge of forwarding the data gathered by the sensors
present within the cluster and the analysis center.
Since the contents of the frames sent out by the
central node of a cluster to the analysis center
originate from multiple body sensors, an intelligent
multiplexing scheme is needed.
4 INTELLIGENT SESSION
MANAGEMENT
This section develops a novel session multiplexing
technique based on wavelet theory. We first discuss
the mathematical aspects related to the wavelet packet
transform. Then, we develop a multiplexing scheme
where data emanating from multiple elementary
sensors can be carried in a unique frame flow. For this
purpose, we introduce a frame structure based on the
parent-child relationship defined in wavelet theory.
The Wavelet Transform (WT) is a time-scale
transform that can be used to perform signal analysis.
It offers effective time-frequency representation of
signals. Wavelet theory and application have matured
in recent decades and have proven to have
tremendous application in fields such as data
compression, multi-scale analysis, transient signal
processing, and more. In practice, the wavelet
transform is implemented using a couple of filters; a
low-pass filter is used to generate approximation
coefficients and a high-pass filter is used to generate
detail coefficients. A decimation phase is also used so
that the size of each of the approximation and detail
signals is half the size of the input signal.
Mallat (Mallat, 1989) showed that a multi-
resolution decomposition of a signal f(t) can be
achieved by iterating the wavelet decomposition on
the approximation signal (which will be initialized to
f(t)). More recently (Feil and Uhl, 1998), a more
sophisticated multi-resolution analysis, based on the
Wavelet Packet Transform (WPT), has been
proposed to apply the wavelet transform to both the
approximation and the detail coefficients at every
decomposition stage. Figure 2 illustrates this
transform where H
0
denotes the low-pass filter and
H
1
denotes the high-pass filter.
Figure 2: Wavelet packet decomposition.
For n levels of decomposition the WPT produces 2n
different sets of coefficients (or nodes) as opposed to
(n + 1) different sets for the DWT. However, due to
the downsampling process the overall number of
coefficients is still the same and there is no
redundancy.
The basic idea of our work is that larger time
slots should be allocated to the sensor nodes that
provide more refined data. For this purpose, we
define a parent–child relationship between wavelet
coefficients, and let the coarser resolution transport
the most refined data. We have investigated
dependencies between wavelet coefficients on this
traffic. As shown in (Feil and Uhl, 1998), the
dependencies between parent-child are very
important. Therefore, the wavelet packet transform
minimizes the cross-correlation between two
decomposed signals. Therefore, a frame issued by an
intermediate sensor node can carry data from
DCNET 2010 - International Conference on Data Communication Networking
88
multiple flows generated by different sensor nodes.
This idea is detailed in the following.
Importance should be attached to the values on
diagonals. The main diagonal is not so important in
our case. More important are other diagonals, which
directly show the dependencies between predecessors
and successors. For example, second diagonal reveals
direct parent-child dependency.
According to Figure 2, if the size of a signal f(t) is
denoted by s, then the size of the signals obtained
after n wavelet stages is s/2i (the rounding operator is
omitted because we suppose that s is a power of 2).
Therefore, if n is the number of elementary sensors
and F is the frame size; then, a fair decomposition of
the frame gives that <F/n> bits are allocated to every
elementary sensor, where <.> denotes the rounding
operator. Hence, the frame can be structured so that
the analysis center reconstructs the signals
corresponding to lower decomposition depths before
those corresponding to deep depths. The role of the
intermediate sensor is simply to increase the depth of
the wavelet packet transform according to the
priority of the corresponding sensor node.
5 SECURITY PROTOCOL
The first step in ensuring secure data aggregation at
intermediate nodes is to enable the intermediate nodes
to have appropriate encryption/decryption keys to
communicate and decipher the incoming data, apply
the aggregation function and relay them forward.
When data is relayed, it is assumed that it is broadcast
(using omni-directional antenna). Furthermore, if the
same data has to be transmitted to several nodes and if
the nodes are operating using distinct pair-wise keys,
then, care must be taken to transmit the data multiple
times, each time encrypted differently with a different
key. This could potentially be a drain on energy and a
hindrance to in-network processing. As we have seen
earlier, having a common key for the group of nodes
is a possible solution to this problem, but has an
inherent weakness in that the whole network could be
compromised if an attacker successfully attacks any
one node.
The basic idea behind our protocol is to make a
sensor independently generate a public key using an
arbitrary string. For example, a sensor collecting data
of type T at time t will first create a string σ =
(sensor_id|t|T). Using this string, the sensor can
derive a public key, π
σ
to encrypt the data and send
them to the storage site. There is no corresponding
secret key created. In fact, the sensor cannot create
the secret key needed to decrypt the message.
When the sensor wishes to release this
information to the analysis center, the sensor can
derive the corresponding secret key, κ
σ
, by using the
same string σ. This secret key only allows the analysis
center to decrypt messages encrypted by a sensor
using the same string. This simplifies key
management, since the sensor can generate the secret
key on-demand without keeping track of which keys
were used to encrypt which data. The only
requirement is that the string used to describe the
event is the same.
Setup: We select an elliptic curve E over GF(p),
where p is a big prime number. We also denote P as
the base point of E and q as the order of P, where q is
also a big prime. A set of n secret keys κ
1
,…, κ
n
GF(q) is chosen to generate the master secret key,
denoted by K = (κ
1
,…, κ
n
). The n public keys are then
generated to make up the master public key, denoted
by Π = (π
1
,…, π
n
), where π
i
= κ
i
.P, 1 i < n. Finally,
a collision resistant one-way hash function is chosen,
The parameters (Π, P, p, q, h(.)) are released as
the system public parameters.
Keygen: To derive a secret key κ
σ
corresponding to a
public key generated by a string σ, the sensor
executes keygen(σ) = κ
σ
,


.
,
where h
i
(σ) is the i-th bit of h(σ).
Encrypt: To encrypt a message m using a public key
derived from string σ, the sensor does encrypt(m,σ) to
determine the ciphertext c.
Algorithm encrypt
Determine string σ using agreed-upon syntax
Generate public key π
σ
where
π
σ
= Pn
i
=1 h
i
(σ) · yi
Execute EccEncrypt(m, π
σ
) to obtain c
Decrypt: The analysis center executes decrypt(c, κ
σ
)
to obtain the original message m which was encrypted
using a secret key derived from σ.
Algorithm decrypt
Requests permission from sensor to obtain data
described by σ
Sensor runs Keygen(σ) to derive κ
σ
Analysis center executes EccDecrypt(c, κ
σ
) to obtain m
Based on these functions, we develop the
following protocols for secure data collection,
transfer, and aggregation.
Secure Data Collection: Having collected an event
d, the sensor executes the following algorithm to
encrypt it.
SECURE WEARABLE AND IMPLANTABLE BODY SENSOR NETWORKS IN HAZARDOUS ENVIRONMENTS
89
Algorithm secure_data_collection
Derive the string σ, and generate a random number n
Calculate m1 = (flag|n) where flag is a known bitstring
Calculate m2 = (d|n)
Calculate c1 =Encrypt(σ,m1)
Calculate c2 =Encrypt(σ,m2)
Secure Data Transfer: Periodically, each sensor in
the WISSN will transfer its data to the analysis center.
This is done by first aggregating all the data into an
intermediate sensor node, which then forwards the
aggregated data to the storage site. Assuming that
there are k tuples generated by the WISSN, the
intermediate sensor will forward the set {(c
11
, c
12
), …
, (c
k
1
, c
k
2
)}.
Secure Data Querying: An analysis center wishing
to obtain data collected under some σ will first
contact the CA for permission. After the CA agrees,
the CA will run Keygen(σ) to derive the
corresponding secret key κ
σ
needed to decrypt data.
Then, the following algorithm is executed to decrypt
the data:
Algorithm Secure data querying
for every (c
i
1
, c
i
2
) i
k for sensor do
Storage site sends ci1 to analysis center
Analysis center runs Decrypt(c
i
1
, σ)
if the initial bits of the result match flag then
Analysis center requests corresponding
c
i
2
from storage site
Analysis center executes Decrypt(c
i
2
, σ)
and checks whether the n matches the
value from c
i
1
Analysis center accepts d if both are correct
end if
end for
Since all the data are encrypted, the storage site
cannot return a specific encrypted tuple to the analysis
center. Instead, the storage site simply lets the
analysis center try to decrypt each tuple (c
1
, c
2
)
belonging to the sensor. The reason for returning c
1
to
the analysis center first instead of returning c
2
directly
is to improve efficiency. Since the length of c
1
is
much shorter than c
2
, letting the analysis center first
attempt to decrypt c
1
before sending the much longer
c
2
reduces transmission time.
The analysis center can check if the data obtained
from the storage site belongs to his sensor by
checking whether the same random number n is used
in both c
1
and c
2
. Since this random n is known only
to the sensor encrypting the data, only that sensor can
embed the same n in both c
1
and c
2
.
6 ASSESSMENT
AND EVALUATION
In this section, we validate the proposed session
management and security protocols. We first analyze
the features of the developed functionalities with
respect to the requirements given in Section III.
Then, we proceed to a performance evaluation based
on simulation of the wavelet-based session
management scheme. Finally, we discuss the security
properties guaranteed by our cryptographic protocols.
A. Proving Features
We discuss the features of the developed
functionalities with respect to real-time
responsiveness, fairness, privacy and anonymity.
Real-time responsiveness: The data acquisition
center is able, via structured queries, to have the data
collected by the sensor nodes nearly in real-time. In
fact, the period between two queries has to be
sufficient for the transmission of n.l.p.s bits, where n
is the number of elementary sensors, l is the event
rate, p is the even gathering periodicity, and s is the
average signal size. For n=10, l=2, p=1mn, and s=2
9
,
we find that the transmission rate between the
elementary sensor and the analysis center should be
approximately 6kb to fulfill the real-time
requirement. The period of time needed to upload all
the collected events to the analysis center would be
0.6 seconds, in that case, using a 10kb/s link.
Fairness: the allocation scheme used when building
the upward frame guarantees an equal slot of time for
all nodes constituting the BAN. Nodes that have
larger quantities of information to send are provided
with greater depth, using wavelet transform, to send
more data in the same period of time. This approach
reduces the latency measured for the arriving data at
the analysis center.
Privacy and Anonymity: privacy provided by a BAN
in a hazardous environment covers personal
information related to the wearer and information
related to the collecting sensors (e.g., used algorithms
and nature of the data collected). After multiplexing
the collected data, the transmitted frame is unable to
show any of the private information since the wavelet
transform will mix these data at variable depths. In
addition, a public encryption is added to this process.
B. Security Evaluation
The main overhead of our protocols is the amount of
time needed to generate a single π
σ
from a string σ
using n number of public keys π
1
,…, π
n
. Note that the
DCNET 2010 - International Conference on Data Communication Networking
90
value of n is not related to the number of different π
σ
s that can be generated. The WISSN can continue to
generate as many π
σ
on-the-fly as needed, regardless
of the value of n. Once π
σ
is generated, the remaining
encryption is the same as that for a regular ECC
encryption.
Figure 3 shows the amount of time needed to
generate a single π
σ
with varying values of n. All n
public keys are initially stored in the flash memory.
Figure 4 shows the amount of flash storage need to
store n different public keys. We see from Figure 3
that, for n = 360, we need only 0.9 seconds to
generate π
σ
.
Figure 3: Time needed to derive one public key versus the
number of elementary keys.
Figure 4: Flash memory storage needed to store one public
key versus the number of elementary keys.
Figure 5 shows the time needed to perform the
encryption once the public key π
σ
has been derived.
For a given piece of data, encrypting with just one π
σ
requires about 1.5 seconds. Again this is the
encryption time for the symmetric key (r), which will
then be used to encrypt the raw data. The symmetric
key can be used for a period, say 10 minutes. The
cost of the 1.5s can be compensated over the 10
minute period. The amount of time needed for
multiple π
σ
s to encrypt the same data is proportional
to the number of π
σ
s. While in Figure 5 the amount
of time needed for 10 different π
σ
is close to 15
seconds while it is worth mentioning that ,in practice,
we are unlikely to use many different public keys to
encrypt the same event.
Figure 5: Time needed for encryption versus number of
keys.
7 CONCLUSIONS
AND PERSPECTIVES
In this paper, we defined an architecture for secure
wearable and implantable smart sensor networks
where an analysis center periodically launches
queries to gather data related to the monitored
environment. To adapt to the hazardous nature of the
contexts where such systems are typically deployed,
we proposed a multiplexing scheme and a
cryptographic protocol based on wavelet packet
decomposition and elliptic curve cryptography
respectively. We have shown that these approaches
provide real-time responsiveness (through intelligent
session management) as well as anonymity (since the
human identity is not involved in the cryptographic
protocol).An extension of the session multiplexing
technique to a physical layer is under development
for use in situations where the optical sensors are
linked to the analysis center via laser beams.
Moreover, a simpler security protocol not involving
intervention by the certification authority is being
developed.
Our future work will also include the design and
deployment of wearable and implantable smart
sensor nodes with light-weight self-abilities to detect
in real time unknown activity patterns, to swiftly
respond to them, and to learn activity patterns over
time and adapt to the dynamism of the hazardous
environment and to changing degree of security and
privacy breaches. Such abilities may enable the
reduction of communication overhead between
SECURE WEARABLE AND IMPLANTABLE BODY SENSOR NETWORKS IN HAZARDOUS ENVIRONMENTS
91
nodes. Applications scenarios such as healthcare and
smart homes will be investigated.
REFERENCES
Robert Puers, “Implantable sensor systems”, DISens
symposium- book 2005.
M. Hamdi, N. Boudriga, “Security In Wireless Sensor
Networks,” Handbook of Research on Wireless
Security, Information Science Reference, ISBN-10:
159904899X, March 2008.
Polar. Polar watches. http://www.polarusa.com.
Suunto. T6, foot pod, n6hr. http://www.suunto.com.
BodyMedia. Healthwear armband, bodybugg. http://
www.bodymedia.com.
K. M. Scannell, D. A. Perednia, and H. M. Kissman.
Telemedicine: Past, present, future. Technical report,
U.S. Department of Health and Human Services.
National Library of Medicine. Reference Section.,
1995.
Martin. T., E. Jovanov, and Raskovic. D. Issues in
wearable computing for medical monitoring
applications: A case study of a wearable ecg
monitoring device. In Proc. Intl. Symp. Wearable
Computers (ISWC’00), pages 43–49, 2000.
Oliver, N. Flores-Mangas, F. HealthGear: a real-time
wearable system for monitoring and analyzing
physiological signals. International Worskhop on
Wearable and Implantable Body Sensor Networks,
2006.
S. Park and S. Jayaraman. Enhancing the quality of life
thourgh wearable technology. IEEE Engineering in
Medicine and Biology Magazine, 22:3:41–48,
2003.Dfd
R. Paradiso. Wearable health care system for vital signs
monitoring. In Proc. IEEE Int. Conf. Information
Technology Applications in Biomedicine, pages 283–
286, 2003.
E. Jovanov, A. Milenkovic, C. Otto, and P. C. de Groen. A
wireless body area network of intelligent motion
sensors for computer assisted physical rehabilitation.
Journal of Neuroengineering and Rehabilitation, 22:6,
2005.
Kemp, J., Gaura, E. I., Brusey, J. 'Instrumenting Bomb
Disposal Suits with Wireless Sensor Networks.' In
Proceedings of the 5th International Conference on
Informatics in Control, Automation and Robotics
(ICINCO 2008). May 11-15 2008 at Funchal, Madeira
– Portugal, 2008.
S. G. Mallat "A Theory for Multiresolution Signal
Decomposition : the Wavelet Representation",. IEEE
PAMI, Vol.11(7), pp.674693, 1989.
M. Feil and A. Uhl. Wavelet Packet Decomposition and
Best Basis Selection on Massively Parallel SIMD
Arrays. Proceedings of the International Conference
“Wavelets and Multiscale Methods” (IWC’98),
Tangier, 1998.
A. Mihovska and N. R. Prasad, Adaptive Security
Architecture based on EC-MQV Algorithm in Personal
Network (PN), Fourth Annual International Conference
on Mobile and Ubiquitous Systems: Networking &
Services (MobiQuitous 2007), Volume, Issue, pp. 1-5,
6-10 August, 2007
W. Leister, H. Abie, A.-K. Groven, T. Fretland and I.
Balasingham, “Threat Assessment of Wireless Patient
Monitoring Systems,” ICTTA'08 (Intl. Conf. on
Information and Communication Technologies: From
Theory to Practice), Damascus, Syria, 7-11 April 08,
2008.
W. Leister, T. Fretland, and I. Balasingham, Security and
Authentication Architecture Using MPEG-21 for
Wireless Patient Monitoring Systems, International
Journal on Advances in Security, 2009 Vol. 2, No. 1
(in press)
DCNET 2010 - International Conference on Data Communication Networking
92