Revisiting Smart Dust with RFID Sensor Networks

Michael Buettner Ben Greenstein Alanson Sample Joshua R. Smith

University of Washington Intel Research Seattle University of Washington Intel Research Seattle
David Wetherall
University of Washington, Intel Research Seattle

A BSTRACT sufficient for the applications.

We argue that sensing and computation platforms that Despite this success, WSNs have fallen short of the
leverage RFID technology can realize “smart-dust” ap- original vision of smart-dust. They have not led to an
plications that have eluded the sensor network commu- approximation of sensing embedded in the fabric of ev-
nity. RFID sensor networks (RSNs), which consist of eryday life, where walls, clothes, products, and personal
RFID readers and RFID sensor nodes (WISPs), extend items are all equipped with networked sensors. For this
RFID to include sensing and bring the advantages of manner of deployment, truly unobtrusive sensing devices
small, inexpensive and long-lived RFID tags to wireless are necessary. The size and finite lifetime of motes make
sensor networks. We describe sample applications suited them unsuitable for these applications.
to the space between existing sensor networks and RFID. We argue in this paper that Radio Frequency Identifi-
We highlight the research challenges in realizing RSNs cation (RFID) technology has a number of key attributes
such as the use of intermittent power and RFID proto- that make it attractive for smart-dust applications. Pas-
cols suited to sensor queries. sive UHF RFID already allows inexpensive tags to be re-
motely powered and interrogated for identifiers and other
1 I NTRODUCTION information at a range of more than 30 feet. The tags can
In the late 1990s, the vision of “smart-dust” was articu- be small as they are powered by the RF signal transmitted
lated by the research community. This vision was pred- from a reader rather than an onboard battery; aside from
icated on advances in microelectronics, wireless com- their paper thin antennas, RFID tags are approximately
munications, and microfabricated (MEMS) sensing that one cubic millimeter in size. Moreover, their lifetime can
were enabling computing platforms of rapidly diminish- be measured in decades as they are reliable and have no
ing size. The early proponents imagined devices one cu- power source which can be exhausted. These advantages
bic millimeter in size with capabilities sufficient to power have resulted in the widespread deployment of RFID for
themselves, sense the environment, perform computa- industrial supply-chain applications such as tracking pal-
tion, and communicate wirelessly [7]. Large-scale de- lets and individual items. However, RFID technology is
ployments of such devices would enable a wide range limited to only identifying and inventorying items in a
of applications such as dense environmental monitoring, given space.
sensor rich home automation and smart environments, The RFID Sensor Networks (RSNs) we advocate in
and self-identification and context awareness for every- this paper extend RFID beyond simple identification to
day objects. in-depth sensing. This combines the advantages of RFID
The past decade has seen significant effort and technology with those of wireless sensor networks. In
progress towards the original motivating applications. In our previous work, we have demonstrated the techni-
particular, wireless sensor networks (WSNs) based on cal feasibility of building small, battery-free devices that
“mote” sensing platforms have been applied to many use the RFID PHY and MAC layer to power themselves,
real-world problems. Remote monitoring applications sense, compute, and communicate; we refer to these de-
have sensed animal behavior and habitat, structural in- vices as Wireless Identification and Sensing Platforms
tegrity of bridges, volcanic activity, and forest fire dan- (WISPs)[12, 13]. While other research efforts such as
ger [5], to name only a few successes. These net- [3] have combined RFID with sensing, to the best of our
works leveraged the relatively small form-factor (ap- knowledge, the Intel WISP is the only RFID sensor node
proximately 1” x 2”) of motes and their multihop wire- with computational capabilities and that operates in the
less communication to provide dense sensing in difficult long range UHF band.
environments. Due to their low power design and care- While the feasibility of WISPs has been established
ful networking protocols these sensor networks had life- by this earlier work, how to harness many such devices
times measured in weeks or months, which was generally to create RSNs is an open question. An RFID sensor net-

1
 mote and its finite lifetime makes it unsuitable for ap-
plications where sensing must be embedded in small ob-
jects, or in inaccessible locations where batteries cannot
be replaced.

2.2 RFID
While there are a number of different RFID specifica-
tions, that of greatest interest for sensing applications
is the EPCglobal Class-1 Generation-2 (C1G2) proto-
col [4], as it is designed for long-range operation. The
C1G2 standard defines communication between RFID
readers and passive tags in the 900 MHz Ultra-High Fre-
Figure 1: Commercial UHF RFID tag, Accelerometer WISP, Telos
mote with batteries quency (UHF) band, and has a maximum range of ap-
proximately 30 feet. A reader transmits information to a
work consists of multiple WISPs and one or more read- tag by modulating an RF signal, and the tag receives both
ers. Consequently, realizing full-scale RSNs will require down-link information and the entirety of its operating
development at both the WISP and the reader, as new energy from this RF signal. For up-link communication,
protocols and techniques must be developed unlike those the reader transmits a continuous RF wave (CW) and the
of either RFID or WSNs. tag modulates the reflection coefficient of its antenna. By
The focus of this paper is the applications that RSNs detecting the variation in the reflected CW, the reader is
enable and the systems challenges that must be overcome able to decode the tag response. This is referred to as
for these to be realized. As the traditional RFID usage “backscattering,” and requires that a tag be within range
model is very different from that of WSNs, RSNs face of a powered reader.
substantial challenges when trying to integrate the two The MAC protocol for C1G2 systems is based on
technologies. For example, unlike WSNs, RSNs must Framed Slotted Aloha [11], where each frame has a num-
cope with intermittent power and unlike RFID must sup- ber of slots and each tag will reply in one randomly se-
port sensor queries rather than simply identification. lected slot per frame. Before beginning a frame, a reader
can transmit a Select command to reduce the number of
2 F ROM M OTES AND RFID TO RSN S active tags; only tags with ID’s (or memory locations)
Two technologies have been widely used to realize real- that match an included bit mask will respond in the sub-
world monitoring applications: wireless sensor networks sequent round. After a tag replies, the reader can choose
via motes, and RFID via standard tags and readers. We to singulate the tag, or communicate with it directly, and
describe and contrast each technology and then present read and write values to tag memory. These mechanisms
their combination (Table 1) as RFID sensor networks enable rapid tag identification and unicast read and write.
(RSNs). We use prior work on the WISP [12, 13] to RFID tags are fixed function devices that typically use
demonstrate the technical feasibility of this combina- a minimal, non-programmable state machine to report a
tion. Representative devices for the three technologies hard-coded ID when energized by a reader. As they are
are show in Figure 1. powered by the reader, the device itself can be very small,
though the antenna requires additional area. As the an-
2.1 Wireless Sensor Networks (Motes) tenna is flexible and paper thin, their small size means
Currently, most WSN research is based on the Telos they can be affixed to virtually any object to be identified
mote [10], which is a battery powered computing plat- However, RFID tags provide no general purpose comput-
form that uses an integrated 802.15.4 radio for commu- ing or sensing capabilities.
nication. These motes are typically programmed to orga-
nize into ad-hoc networks [15] and transmit sensor data
2.3 RFID sensor networks (WISPs + readers)
across multiple hops to a collection point. To extend We define RFID sensor networks (RSNs) to consist
network lifetime, motes duty cycle their CPU and radio of small, RFID-based sensing and computing devices
(e.g., with low-power listening [9]), waking up intermit- (WISPs), and RFID readers that are part of the infras-
tently to sense and communicate. With a duty cycle of tructure and provide operating power. RSNs bring the
1%, motes can have a lifetime of up to three years before advantages of RFID technology to wireless sensor net-
the batteries are exhausted. works. While we do not expect them to replace WSNs
Using multihop communication, WSNs can sense over for all applications, they do open up new application
great distances, which has made them idea for a wide spaces where small form-factor, long-lived, or inacces-
range of applications. However, the large size of the sible devices are paramount. Our hope is that they will

2
 CPU Sensing Communication Range Power Lifetime Size (inches)
WSN (Mote) Yes Yes peer-to-peer Any battery < 3 yrs 3.0 x 1.3 x .82 (2.16 in3 )
RFID tag No No asymmetric 30 ft harvested indefinite 6.1 x 0.7 x .02 (.08 in3 )
RSN (WISP) Yes Yes asymmetric 10 ft harvested indefinite 5.5 x 0.5 x .10 (.60 in3 )

Table 1: Comparison of Technologies

elegantly solve many sensor network applications, e.g., possible to gauge exposure during transport from a donor
home sensing and factory automation where installing or to a bank, between banks, and ultimately to a patient.
carrying readers is feasible. WISPs with temperature sensors could be attached di-
Prior work at Intel Research demonstrates that WISPs rectly to individual blood bags and queried for their mea-
can be built today. The most recent Intel WISP is a wire- surements. Such sensors must be small (one could imag-
less, battery-free platform for sensing and computation ine affixing sensors with something like a price tag gun),
that is powered and read by a standards-compliant UHF and inexpensive to the point of being disposable.
RFID reader at a range of up to 10 feet. It features a To understand the challenges in building such an ap-
wireless power supply, bidirectional UHF communica- plication, an Intel WISP was attached to a container of
tion with backscatter uplink, and a fully programmable milk (a suitable and widely available approximation of
ultra-low-power 16-bit flash microcontroller with analog a bag of blood), and its temperature was monitored over
to digital converter. This WISP includes 32K of flash the course of 24 hours [16]. For this study, a storage
program space, an accelerometer, temperature sensor, capacitor (roughly the size of a pea) was attached to the
and 8K serial flash. Small header pins expose micro- WISP to log sensor data for up to a day when out of range
controller ports for expansion daughter boards, external of a reader.
sensors and peripherals.
The Intel WISP has been used to implement a variety 3.2 Brains
of demonstration applications that read data from a sin- Research in neuroscience has explored using neural sen-
gle sensor unit. These include the first accelerometer to sors for controlling prosthetic limbs [14]. Sensors placed
be powered and read wirelessly in the UHF band, and outside the skull can capture neural activity, but the sig-
also the first UHF powered-and-read strain gage [17]. nals are too coarse-grained and noisy to be effective.
Even without its sensing capabilities, the Intel WISP can With surgery, sensors can be placed directly on the brain,
be used as an open and programmable RFID tag: the resulting in much higher resolution and finer control of
RC5 encryption algorithm was implemented on the In- the limb. Using conventional technologies (e.g., motes)
tel WISP [2]. We believe this is the first implementation presents difficulties with respect to power because batter-
of a strong cryptographic algorithm on a UHF tag. ies need to be replaced via invasive surgical procedures,
as is the case with pacemakers.
3 E XAMPLE A PPLICATIONS An RFID sensor network is well suited to this applica-
tion. A patient would have WISPs equipped with neural
RFID sensor networks have broad applicability wherever probes placed inside the skull. These could then draw
sensing, small form factor, embeddability, longevity, and power from and communicate with a device outside the
low maintenance are desired and fixed or mobile readers body, e.g., an RFID reader worn as a cap, bracelet, or
are feasible. This section highlights applications within belt. We have completed initial studies that show the fea-
this space and some of the key design considerations. sibility of integrating neural sensors with the WISP [6].
3.1 Blood 3.3 The Elderly
Blood transfusions save lives, replacing blood lost dur- Providing care for the elderly is one of the largest health-
ing surgery, illness, or trauma. After donation, blood is care costs facing us today, particularly as the “baby
bagged and refrigerated between 1◦ and 6◦ C and has a boomer” generation ages. Keeping people in their homes
shelf life of about 35 to 42 days. Refrigerators used to for as long as possible significantly reduces these costs
store blood are monitored for outages and temperature and increases quality of life. The difficulty with this is
fluctuations, and collection dates are recorded on blood in detecting and reacting to emergencies, such as the pa-
bags. However, the temperature of the bag itself is rarely tient falling or forgetting to take critical medication. Cur-
monitored with any regularity. This makes it difficult to rently, families have no choice but to hire costly support
determine if a given bag was warmed to unsafe levels, personnel to regularly check-in on their loved ones.
such as if it is near the front of the refrigerator and the Traditional RFID has been explored to help monitor
door is often opened. Additionally, it is difficult or im- the behavior of the elderly. For example, by having the

3
patient wear a short range RFID reader bracelet and plac- tion, and reading/writing to memory all consume differ-
ing RFID tags on a toothbrush, toothpaste, and faucet, ent amounts of energy.
software can infer that an elderly person is brushing her To run tasks to completion, WISPs will require sup-
teeth when these tags are read in succession [8]. Such port for intermittently powered operation. They must be
fine-grained sensing requires very small devices, and is able to estimate the energy required to complete a task,
simpler and more respecting of privacy than competing perhaps based on task profiling or energy budgets, and
approaches using computer vision, where video of the compare it with estimated reserves. To work well in this
person is continuously recorded and analyzed. regime, RSN devices may also need to cooperate with
Adding sensing (e.g., an accelerometer) to long range RFID readers for power management. This would in-
RFID tags would have several key advantages. Rather volve signaling by either the reader, of its intended trans-
than requiring a person to wear a short-range reader, mission time, or by the WISP, of its needs. Even with
which can be taken off, a few long-range readers could signaling, it will be difficult to predict power expecta-
be placed in the home and behavior could be determined tions because the rate at which energy is harvested de-
via direct communication with the objects that are being pends on the frequency of the reader and the proximity
interacted with. This explicit information would be more of the device to the reader, both of which will change
accurate in detecting behavior than inference based only over time. Thus, to increase the kinds of tasks that could
on object identifiers. be supported, large tasks might need to be split into
RSNs are an appropriate solution for the above appli- smaller, restartable stages; intermediate results between
cations and those like them. Our initial studies using the the stages could be maintained in device storage (flash or
WISP show the potential of existing RFID sensing de- RAM) or be offloaded to the reader.
vices for use in such applications. However, these stud- To extend functionality when away from a reader, one
ies involved only a single WISP. Combining many such approach would be to provide a small amount of energy
devices into a full RSN will require further research. storage on the device, e.g., a capacitor, and store excess
energy when close to an active reader. This storage ca-
4 C HALLENGES pacitor would be small relative to a battery, because it
would be intended only for short term usage and is wire-
RSNs combine the technology of RFID and sensing with lessly recharged over time. The Data Logger WISP used
the usage models of sensor networks. At the device level, for the milk carton study takes this approach, using a
the WISP shows that it is feasible to combine sensing super-capacitor that, when fully charged, sustains low
with RFID. However, at the systems level, challenges duty-cycle operation for more than a day. The type of
arise due to the mismatch between the RFID usage model tasks that this WISP enables would be limited, due to en-
and that of wireless sensor networks. We detail several ergy requirements, and the period of functionality would
challenges in this section. be limited due to leakage. Additionally, unpowered op-
eration would likely stress tradeoffs between stages. For
4.1 Intermittent Power example, writing to flash is significantly more energy in-
RFID readers provide an unpredictable and intermittent tensive than computing with RAM but preserves valuable
source of power. This makes it difficult for WISPs to as- data for later use.
sure that RSN tasks will be run to completion. WISPs
are powered only when in range of a transmitting RFID 4.2 Asymmetric Sensing Protocols
reader and, for regulatory and other reasons, readers do The communication paradigm of RFID results in systems
not transmit a signal continuously. Instead, they trans- that are limited by up-link bandwidth. When the data
mit power for a brief period before changing channels or of interest is simply each tag’s identity, this constraint
entirely powering down. For standard RFID tags where is not a problem. However, it makes it difficult to de-
the task is simply to transmit the identifier, this style of velop efficient protocols for gathering sensor data that
communication is sufficient. However, it is a poor fit for changes over time. In WSNs, nodes are peers in terms
RSN tasks that span many RFID commands. of the physical and link layers of their communication,
The WISP harvests energy from a reader and stores e.g., each mote has an 802.15.4 radio capable of send-
this energy in a capacitor. When enough energy is har- ing and receiving transmissions with other nodes that are
vested, the WISP powers up and can begin sensing and in range. In contrast, because they draw on RFID, RSN
communicating. However, sensing and communication nodes are highly asymmetric in terms of their commu-
drain power from the WISP. This can result in the WISP nication abilities. With RFID, readers are able to trans-
losing power in the middle of an operation depending on mit messages to all tags and tags can transmit messages
the task and the reader behavior. A further complica- to the reader. However, tags can do so only when the
tion is that receiving, transmitting, performing computa- reader initiates communication, and tags cannot commu-

4
nicate directly with each other even when powered by the ever, the communication model of passive RFID means
reader. that the reader can have a large degree of control over
These differences complicate the design of protocols what code is loaded and executed on the WISPs at any
for gathering sensor data. Currently, WISPs with new point in time.
sensor data must wait until they are interrogated by a In contrast to mote message reception, which con-
reader. This increases the likelihood of many devices sumes energy from small onboard batteries, WISP mes-
wanting to use the bandwidth limited channel at the same sage reception is powered entirely by the reader. Thus,
time. Techniques to perform data pre-processing within frequent and large code transfers are feasible, which
the network (on each RSN device) could help to some would allow for the complete retasking of WISPs with
extent. However, the standard RFID strategy of identify- costs in terms of latency only. Moreover, since down-
ing and then communicating with each device is wasteful link communication is cheap when in range of the reader,
as only some devices would have relevant data – a more WISPs might not need to be as “smart” as motes. Rather
dynamic strategy based on the value of the sensor data than requiring WISPs to interpret queries, readers could
would be more effective. tell WISPs exactly what to do, down to the instruction
Consider the eldercare application. A reader might level.
have hundreds of accelerometer WISPs in its field of To fully exploit the potential of RSNs, new tools must
view. Because all the WISPs share a single reader chan- be developed. As a first step, we are developing an RFID
nel, the update rate per tag would be very low if every tag reader platform based on the Universal Software Radio
were simply queried for sensor data sequentially. How- Peripheral (USRP). This platform, when used in con-
ever, at any given moment, only a few objects would typ- junction with the WISP, would allow for the development
ically be in motion (and therefore producing non-trivial of new protocols at both the MAC and PHY layers. Thus
accelerometer sensor values). Furthermore, the set of ob- far we have used it for monitoring RFID systems [1].
jects that are moving would change dynamically, as ob-
jects are put down and picked up. One might want a pro- 4.3 Repurposing C1G2
tocol which gives priority to the most active objects, po- There would be substantial practical benefit to realizing
litely “yielding” to new objects when they start to move. RSN protocols using the primitives of the C1G2 stan-
Existing RFID solutions do not support anything like dard: Commercial off-the-shelf readers could be used for
this functionality. As a first step, one could have WISPs RSN research and deployment, and WISPs would inter-
with sensor activity below a threshold not respond to operate with ordinary (non-sensing) tags. However, the
the reader. But an appropriate threshold level might de- extent to which RSN protocols could be implemented
pend on what is occurring in the room, and such a sim- within the C1G2 standard is an open research question.
ple scheme would not support the “polite yielding” de- Additionally, there is the practical consideration of com-
scribed above. mercial readers not exposing low-level functionality and
For another example of what RSN protocols might be not implementing the complete C1G2 specification. Be-
asked to do, consider the blood application. When many cause of this, even RSN protocols built on top of the
blood bags are read simultaneously, one might want to C1G2 specification might not be implementable using
interrogate the bags with the largest temperature excur- standard readers.
sions first. But since the distribution of temperature ex- Our experience with the Intel WISP suggests that basic
cursions would not be known a priori by the reader, the RSN applications could be approximated using standard
protocol would need to (implicitly) estimate this infor- C1G2 readers. To read sensor data from a C1G2 WISP,
mation. It might for example ask if any WISP has a larger the device would first be singulated, at which point a
temperature excursion than E. If no device responds, the temporary handle would be requested from the tag. A
E response threshold could be repeatedly halved until the reader could then use this handle to address the device
appropriate scale is found. The key requirement would and read sensor data from pre-defined memory locations.
be to estimate an aggregate property of the data without However, the handle would persist only until the reader
exhaustively collecting that data. Finally, RSN protocols singulates another tag or the tag loses power. Thus, read-
might be power aware as well. A WISP that was about to ing from more than one WISP would incur substantial
lose power might be given priority over those with ample protocol overhead due to singulation and handle man-
power. agement. Consequently, simple use of the existing C1G2
As the WISP has limited program space, it may not protocol could provide some level of sensing functional-
be possible to program a WISP such that it will be well ity, but at a significant cost in terms of efficiency.
matched to all possible application scenarios. For exam- Along with reading sensor data, the C1G2 protocol
ple, WISPs may need different protocols for different de- could support basic sensor queries using the Select com-
ployments, and these needs may change over time. How- mand. If the reader knows that a sensor value is written

5
to a particular memory location, it could issue a Select for monitoring weather conditions in wildland fire
command with a mask which matches that location for environments. In Proc. MobiSys, 2006.
sensor values over a given threshold. Consequently, only
[6] J. Holleman, D. Yeager, R. Prasad, J. Smith, and
WISPs with sensor values over that threshold would re-
B. Otis. Neural wisp: An energy harvesting wire-
ply during the next frame. More generally, the Select
less brain interface with 2m range and 3uvrms input
command could be used as a general purpose broadcast
referred noise. In Proceedings of the 31th European
channel. The bit mask in the command could be repur-
Solid-State Circuits Conference, Grenoble, 2005.
posed and interpreted, in the most general case, as op-
codes and data. As multiple Selects could be sent be- [7] J. Kahn, R. Katz, and K. Pister. Emerging chal-
fore each frame, complex tasking and querying could be lenges: Mobile networking for ‘smart dust. Jour-
achieved in this manner. nal of Communication Networks, pages 188–196,
The above mechanisms show that there is potential 2000.
for using the C1G2 standard to implement RSN pro-
tocols. This would have the advantage of being im- [8] M. Philipose, K. Fishkin, M. Perkowitz, D. Patter-
plementable using current reader technology, given a son, D. Fox, H. Kautz, and D. Haehnel. Inferring
reader that is sufficiently programmable. However, these activities from interactions with objects. IEEE Per-
mechanisms may prove too inefficient or may simply be vasive Computing, 3(4):50–57, 2004.
poorly matched to many applications. Further experi- [9] J. Polastre, J. Hill, and D. Culler. Versatile low
mentation is needed. power media access for wireless sensor networks.
In SenSys, 2004.
5 C ONCLUSION
By exploiting RFID technology, we believe that we can [10] J. Polastre, R. Szewczyk, and D. Culler. Telos:
expand the application space of wireless sensor net- Enabling Ultra-Low Power Wireless Research. In
works to ubiquitous, embedded sensing tasks. We have Proc. IPSN/SPOTS, 2005.
sketched sample sensor network applications in the space [11] L. G. Roberts. Aloha packet system with and with-
between traditional mote networks and RFID for supply- out slots and capture. SIGCOMM Comput. Com-
chain monitoring. We have described key systems and mun. Rev., 5(2):28–42, 1975.
networking challenges related to intermittent power and
RSN protocols for sensor queries. We expect RSNs to be [12] A. P. Sample, D. J. Yeager, P. S. Powledge, and
a fruitful new space for networking and systems research, J. R. Smith. Design of an rfid-based battery-free
as there is significant work that must be done to translate programmable sensing platform. In IEEE Trans-
the capabilities of the WISP into full-fledged RSNs. actions on Instrumentation and Measurement (ac-
cepted), 2008.
R EFERENCES
[13] J. R. Smith, A. P. Sample, P. Powledge, A. Mami-
[1] M. Buettner and D. Wetherall. An empirical study shev, and S. Roy. A wirelessly powered platform
of uhf rfid performance. In Proc. Mobicom (to ap- for sensing and computation. In Proc. Ubicomp,
pear), 2008. 2006.
[2] H. J. Chae, D. J. Yeager, J. R. Smith, and K. Fu. [14] D. Taylor, S. H. Tillery, and A. Schwartz. Direct
Maximalist cryptography and computation on the cortical control of 3d neuroprosthetic devices. Sci-
wisp uhf rfid tag. In Proc. Conference on RFID ence, 296:1829–1832, 2002.
Security, 2007.
[15] A. Woo, T. Tong, and D. Culler. Taming the un-
[3] N. Cho, S. Song, S. Kim, S. Kim, and H. Yoo. A derlying challenges of reliable multihop routing in
5.1-uw uhf rfid tag chip integrated with sensors for sensor networks. In Proc. SenSys, 2003.
wireless environmental monitoring. In IEEE Bio-
logical Circuits and Systems (BioCAS) 2008, sub- [16] D. Yeager, R. Prasad, D. Wetherall, P. Powledge,
mitted, 2008. and J. Smith. Wirelessly-charged uhf tags for sen-
sor data collection. In Proc. IEEE RFID, 2008.
[4] EPCglobal. EPC radio-frequency identity protocols
class-1 generation-2 UHF RFID protocol for com- [17] D. J. Yeager, A. P. Sample, and J. R. Smith. Wisp:
munications at 860 mhz-960 mhz version 1.0.9. A passively powered uhf rfid tag with sensing and
2005. computation. In M. I. Syed A. Ahson, editor, RFID
Handbook: Applications, Technology, Security, and
[5] C. Hartung, R. Han, C. Seielstad, and S. Holbrook. Privacy. CRC Press, 2008.
Firewxnet: a multi-tiered portable wireless system