building the internet of things using rfid the rfid ecosystem experience

7/27/2019 Building the Internet of Things Using RFID the RFID Ecosystem Experience

http://slidepdf.com/reader/full/building-the-internet-of-things-using-rfid-the-rfid-ecosystem-experience 1/8

48 Published by the IEEE Computer Society 1089-7801/09/$25.00 © 2009 IEEE IEEE INTERNET COMPUTING

I n t

e r n e

t o

f T h

i n g s

T r a c

kEdi tors : Frédér ic Thiesse • f reder ic . th [email protected]

Flo r i an Michahe l l e s • fmichahel [email protected]

The rapid proli eration o passiveRFID tags in the past decade hasgiven rise to various concepts that

integrate the physical world with the virtual one. One o the most popular isthe Internet o Things (IoT), a vision in

which the Internet extends into our ev-eryday lives through a wireless network o uniquely identi able objects. Givennumerous predictions that we’ll havehundreds o billions o RFID-tagged ob-

jects at approximately ve cents per tagby 2015, 1 we’re not only approachingsuch a world, we’re on its doorstep.

In this type o RFID system, eachphysical object is accompanied by arich, globally accessible virtual objectthat contains both current and histori-

cal in ormation on that object’s physi-

cal properties, origin, ownership, andsensory context ( or example, the tem-perature at which a milk car ton is beingstored). When ubiquitous and availablein real time, this in ormation can dra-matically streamline how we manu ac-

ture, distribute, manage, and recycleour goods. It can also trans orm theway we per orm everyday activities by giving applications current and detailedknowledge about physical events. This“real-li e” context can unlock the door to various business, environmental,personal, and social contexts hithertoinaccessible to Internet applications.

The incredible amount o in orma-tion captured by a trillion RFID tagswill have a tremendous impact on our

lives. However, questions remain i we

At the University of Washington, the RFID Ecosystem creates a microcosmfor the Internet of Things. The authors developed a suite of Web-based,user-level tools and applications designed to empower users by facilitatingtheir understanding, management, and control of personal RFID data andprivacy settings. They deployed these applications in the RFID Ecosystem andconducted a four-week user study to measure trends in adoption and utilizationof the tools and applications as well as users’ qualitative reactions.

Evan Welbourne,Leilani Battle, Garret Cole,Kayla Gould, Kyle Rector,Samuel Raymer,Magdalena Balazinska,and Gaetano Borriello

University of Washington

Building the Internetof Things Using RFIDThe RFID Ecosystem Experience



MAY/JUNE 2009 49

Building the Internet of Things Using RFID

are to use RFID in the IoT. How do we trans ormlow-level RFID data into meaning ul, high-levelin ormation? Can we design and build applica-tions that are tru ly use ul and not just novelties?I so, will their utility outweigh the potentialloss o privacy, and how can we help users un-derstand and control their privacy settings?

At the Universit y o Washington, we’reexploring these issues rst-hand with abuilding- scale, community-oriented researchin rastructure called the RFID Ecosystem(http://r d.cs.washington.edu). This in rastruc-ture creates a microcosm or the IoT in whichwe can investigate applications, systems, andsocial issues that are likely to emerge in a re-alistic, day-to-day setting. We’ve developed asuite o Web-based, user-level tools and appli-cations or the IoT and deployed it in the RFIDEcosystem. We’ve also conducted a our-week user study to investigate patterns o adoptionand utilization o our tools and applications aswell as users’ subjective reactions. We presentthe results o this study, ocusing on tool andapplication usage.

The RFID Ecosystem We built the RFID Ecosystem around an Elec-tronic Product Code (EPC) Class-1 Generation-2

RFID deployment that spans all seven foors o our 8,000-square-meter computer science andengineering building (see Figure 1). The deploy-ment includes 44 RFID readers (each equippedwith up to our antennas or a total o 161) posi-tioned at the building’s entrances, on the stair-wells, and throughout the corridors. Readersrun embedded Linux and have wired or wirelessGigabit Ethernet over which they report their RFID data to a central server. Volunteers carry RFID tags as badges and attach tags to personalobjects. Because most everyday objects aren’t

yet manu actured with tags embedded, and be-cause manu acturer-assigned metadata mightnot be personally meaning ul, we create thetag-object association manually. For this pur-pose, we created a special kiosk where users canselect an RFID tag, physically attach it to anobject, and create a corresponding associationbetween the tag and that object.

All readers in our deployment run customso tware that processes new RFID data be orestreaming it to the central server. This so t-ware continuously polls the reader hardware

or newly detected RFID tags and generates

one tag-read event (TRE) per tag per antenna

per second, a tuple with the schema (tag ID,antenna ID, time) . For example, i tag A isdetected by reader antenna X at time stamp t ,then the custom reader so tware will generateand send the ollowing TRE to the server: (tagA, antenna X, t) . Each reader also runs thenetwork time protocol to synchronize its clock with the rest o the system.

We store all TRE data in a central SQLServer database (www.microso t.com/SQL). Thisdatabase also contains metadata about the de-ployment, including each antenna’s latitude

and longitude and a symbolic antenna name( or example, “ ront entrance,” “4th foor stair-well,” or “Room CSE 405”). We wrote so twareto transmit data between the readers and thedatabase in Java using Apache’s MultipurposeIn rastructure or Network Applications library

or e cient, secure networking. This so twareimplements various privacy policies 2,3 and runsthe Cascadia system 4 to support application de-

velopment and execution. Finally, our tools andapplications are entirely Web-based, imple-mented with the Google Web Toolkit (http://

code.google.com/webtoolkit/), and hosted with

Figure 1. The RFID Ecosystem. RFID reader antennas are mounted on cable trays (upper left) and in custom-built wooden boxes(lower left). An RFID kiosk (upper right) lets users associate one of three types of tags (lower right) with a personal object.



Internet of Things Track

50 www.computer.org/internet/ IEEE INTERNET COMPUTING

Apache and Tomcat (www.apache.org) on a sep-arate server.

User-Level ToolsThe RFID data our deployment supplies o er low-level location in ormation in terms o tagsand antennas. TREs — such as (tag A, an-tenna X, t) or (tag B, antenna Y, t +1) — are help ul to IoT users only i a middle-ware can trans orm them into more meaning-

ul, high-level in ormation about events thatapplications and users can directly consume( or example, Ana is leaving the o ce with her purse). Furthermore, because such high-level

events are personal and potentially sensitive,users must be able to precisely control all in-ormation disclosure to avoid privacy breaches.

As such , we’ve developed several secure, Web-based tools that let users directly control howtheir RFID data is trans ormed and disclosed inthe RFID Ecosystem.

Transforming Low-Level RFID DataTo support trans orming TREs into higher- levelevents, we built tools that let users direct ly de-

ne metadata and associate it with tags and

antennas. One such tool, the Tag Manager,

presents a highly interactive set o menus, ta-bles, and Web orms or creating and managingmetadata on a user’s tags and personal objects.The Tag Manager inter aces with the RFID ki-osk so that when users are at the kiosk, they can associate one or more physical tags with anobject. For example, a new user can use the TagManager at the kiosk to register several per-sonal tags. Later, the same user can access theTag Manager rom his laptop to delete objectsand review or edit object metadata (such asname, type, image URL, or where the object’stags were last seen).

A second tool, the Place Manager, suppor tscreating and editing high-level location in-

ormation items, called places .5 A place in theRFID Ecosystem is a set o one or more RFIDantennas with a label. For example, the two an-tennas in the corridor on either side o a user’so ce door might be grouped and labeled “my o ce.” The Place Manager displays each RFIDantenna’s location as an icon in a Google Mapmashup o the RFID Ecosystem deployment. Us-ers can create or edit a place by clicking on an-tenna icons to select or deselect antennas andby entering the place label in a text box (seeFigures 2a and 2b).

Once the Tag and Place Managers de ne

metadata that binds tags to objects and anten-nas to places, respectively, applications andother system components can use that data togenerate higher-level in ormation that’s person-alized and more directly meaning ul to users.

An additional third tool, Scenic, lets users spec-i y what higher-level events they would liketo have extracted rom their TREs (see Figure2c). Scenic uses an iconic visual language anda storyboard metaphor to describe how peopleand objects enact an event through a sequenceo movements between places. Speci cally, the

Scenic inter ace lets users drag and drop iconsrepresenting people, objects, places, and basicrelationships (such as inside, outside, near, or

ar) onto one or more panels, or “scenes,” in astoryboard to speci y a movement sequencecorresponding to an event. Thus, to speci y anevent, users simply “tell the event’s story,” sceneby scene. For instance, a user could drag iconsrepresenting himsel , his keys, and a lab intoone scene with icons representing the “with”and “outside” relationships to indicate that he’soutside the database lab with his keys. Then, he

creates a second scene with the same icons, re-

Creates object: keys Creates place: database laboratory

Creates event: “Evan enters database laboratory with keys”

(a) (b)

(c)

Figure 2. Metadata management tools. (a) The Tag Manager creates and manages virtual objects to which tags can be bound;(b) the Place Manager groups antennas into places; and (c) Scenic uses objects and places to specify how to transform a user’s tag-read events (TREs) into higher-level events.



MAY/JUNE 2009 51


placing “outside” with “inside” to indicate thathe’s entered the lab with his keys. Further de-tails on Scenic’s inter ace and implementationare available elsewhere. 4

Controlling PrivacyRFID security and privacy present many chal-lenges, and potential solutions, rom hardwareand wireless protocol security to the manage-ment, regulation, and sharing o collected RFIDdata. 6,7 Our privacy work in the RFID Ecosystemhas ocused on controlling access to collectedRFID data. 2,3 As such, we accomplish privacy control in the ecosystem chiefy through per-sonal data auditing and by en orcing novel ac-cess-control policies. Two tools let users direct ly interact with their personal RFID data and withthe access-control ramework that governs datadisclosure. Both can operate in conjunction withour physical access control (PAC) policy. We dis-cuss PAC in other work, 2 but the main idea isthat users’ access is constrained to events thatoccur red only when and where they were physi-cally present. In this way, their personal RFIDdata can serve as a detailed log o events they might have observed in person throughout theday. Whether a system employs PAC or not, us-ers can review or delete their data with the Data

Browser, or extend the data they choose to sharewith the Access Control Inter ace (see Figure 3).

The Data Browser lets users review all TREscollected on their tags in an interactive, table-based inter ace. The table displays only TREsthat occur in a user-speci ed time window (thede ault window is the current day so ar). Eachtable row shows human-readable in ormationabout an individual TRE, including the taggedobject’s name, the reading antenna’s symboliclocation, and the time stamp. The Data Browser sorts rows in chronological order, but users can

also sort based on object or location by click-ing that eld in the header row. Users can de-lete their data by selecting and deleting speci crows. To make large deletions easier, the DataBrowser has a deletion menu with which userscan quickly delete all data collected over thepast 30 minutes, an entire day, or some other user-speci ed time range.

In addition to directly managing their per-sonal RFID data, users can use the Access Con-trol Inter ace to control what data the RFIDEcosystem automatically stores and discloses

about their tags. This inter ace eatures a set

o Web orms that let users extend the data setthey share with others by de ning additionalcircumstances in which the system can disclosetheir data. For example, a user could de ne arule that says “Pro essors can see when I’m inmy o ce during business hours” or “My riends

can see my location at any time.” Speci cally,the inter ace lets users de ne riend groups towhich they can apply various access-controlrules concerning places and time ranges. Thus,all people in a particular group have accessto the data that PAC (i it’s in use) or a user’saccess- control rules allow. The key advantageto this type o access control is that it increasesthe amount o data available to applications in away that’s tied to physical events, a method us-ers can more easily understand.

RFID-Based Web ApplicationsTo illustrate what’s possible in the IoT withRFID and our user-level tools, we developedand deployed several Web-based applications.These applications combine TREs with metada-ta on objects, places, and events in accordancewith user-de ned access-control rules to o er services to users. The applications we presentrely chiefy on place-level location in orma-tion, unlike other recent RFID applications thatuse more ne-grained locations or in ormationabout people’s and objects’ close-range interac-

tion.8,9

Although several location-based Web

Figure 3. The Data Browser and Access Control Interface. Withthese tools, users can directly interact with their personal RFID dataand with the access-control framework that governs data disclosure.





in rastructures and applications are emerging(see http:// reeagle.yahoo.net or http://plazes.com), we eel that RFID-enabled in ormation on

indoor location, object location, and events willprovide signi cant incremental unctionality.Figure 4 shows our applications.

A Search Engine for ThingsPerhaps the simplest Web-based RFID applica-tion is a search engine or things. We imple-mented a Web inter ace that lets users view thelast recorded location or their tagged objectsor search or a particular object’s location. TheEvent Noti er, a more proactive extension o this application, leverages user-de ned events

to noti y users when the last recorded objectlocation matches some condition. For example,the application can send an SMS reminder tousers when they leave the building without aparticular item that they’ve designated as im-portant (such as a laptop power cable).

Social ApplicationsSome o the most popular Web services o er in ormation and updates on activities in our social networks; eeds rom sites such as Twit-ter (http://twitter.com) and Plazes are some

examples. To explore this space, we developed

an application called R dder that uses eventsabout people and places to give users real-timeupdates in their social networks. Users’ R d-der inter aces display a eed o events thattheir riends have de ned (“Fiona has enteredher o ce,” or “Raj is taking a co ee break”).Users can control their riend lists as well aswhat events are disclosed to which riends by de ning access-control rules with the AccessControl Inter ace. R dder also integrates withTwitter to provide greater utility and socialnetworking capability.

Personal TrendsHistorical queries about object and event datalet users study trends in their activities over time. This can be extremely use ul or applica-tions that support long-term activities such asbusiness projects and collaborations. We built aDigital Diary application that records and dis-plays events in a Google Calendar (http://code.google.com/apis/calendar) or later perusal. Thisway, users can look back over their diaries tosee how and with whom they’ve spent their time. We’ve also added support or plotting his-torical trends using the Google Charts API. Us-ers’ charts can succinctly display where, how,and with whom or what they’ve spent their time

over some arbitrary period. For example, Fionacan have her diary record how o ten she entersthe building with her bicycle helmet and later use that data to visua lize how o ten she biked towork during the past month or year.

Event-Based Desktop SearchThe log o events that applications such as theDigital Diary collect can also enable search-based applications that leverage a user’s memory o events in the physical world. To demonstratethis, we implemented an event-based search

plug-in or Google Desktop Search (http://desk top.google.com). Our plug-in retrieves digitaldocuments created or modi ed around the timeo some physical event that the user remembers.I Ana remembers that she visited a Web siteduring a meeting with her advisor a ew weeksago but doesn’t remember exact ly what that sitewas, she can search by the event “advisor meet-ing” to select all Web sites visited during meet-ings with her advisor.

Usage Patterns

To better understand how people might use our

Figure 4. RFID Web applications. We developed the R dder, Event Noti er, Object Search,Digital Diary, and Event-Based Desktop Searchapplications.



MAY/JUNE 2009 53


tools and applications in the IoT, we conducted aour-week user study with the RFID Ecosystem.

We recruited 67 participants, including 33 un-dergraduates, 30 graduates, two aculty mem-bers, and two sta who occupy the computer science building on a daily basis. We o eredparticipants US$50 as an additional incentiveto use our applications. Each participant worea tag as a personal badge and attached tags to

personal objects using the Tag Manager withthe RFID kiosk. Participants tagged a total o 324 objects in 19 di erent categories (such aswallets, bags, jackets, mugs, books, power ca-bles, and laptops). Further details on the study,including measurements on data rates and vol-umes, tag-read rates, and system per ormance,are available elsewhere. 10

Throughout the our weeks, participants car-ried their tags every day and had access to theTag Manager, Place Manager, Data Browser, and

Access Control Inter ace as well as the Object

Search and R dder applications. Note that we

integrated the Object Search application withthe Tag Manager tool in this study to providea centralized point o access or object-relatedin ormation. In addition to RFID data, regular survey eedback, and a detailed exit survey, wecollected detailed logs o each participant’s in-teraction with tools and applications. Figure 5provides a high-level summary o how partici-pants interacted with their RFID data. In par-

ticular, the plots show operation counts, wherean operation represents one database query (selection, insert, update, deletion, and so on) aparticipant conducted on his or her data. Par-ticipants per orm operations implicitly wheninteracting with any tool or application.

The rst plot shows the cumulative distri-bution unction (CDF) o the total operationcount over all 67 participants. The CDF indi-cates that roughly 40 percent o participantsper ormed ewer than 100 operations, whichrefects minimal system use. In the exit sur-

vey, these users explained that the system

500 1,000 1,500 2,000

P a r t i c i p a n t s

( % )

O p e r a t i o n c o u n t

O p e r a t i o n c o u n t

T o t a l o p e r t a i o n

( % )

Total operation count Day

7 14 21 28

Staff FacultyGradUndergrad

0

20

40

60

80

100

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

0

10

20

30

40

50

60

Staff FacultyGradUndergrad

0

500

1,000

1,500

2,000

2,500

Data BrowserTag ManagerR dderAccess Control InterfacePlace Manager

DataBrowser

Tag Manager,object search

R dder AccessControlInterface

PlaceManager

Objectlocationqueries

Personlocationqueries

Creatingmetadata

Updatingmetadata

Deletingmetadata

Deletingtag-readevents

Modifyingaccess

control

(a) (b)

(c) (d)

592

3,794

2,601

18426

2,302

1,026

417

13820 1

184

Figure 5. Results of the four-week user study. We measured (a) a cumulative distribution function (CDF) of operationsper participant, (b) tool and application usage over time, (c) the number of operations performed through each tool

and application, and (d) how many times participants performed each type of operation.





wasn’t detecting t heir tags well enough or theapplications to be use ul. The other 60 per-cent o users didn’t have such severe problems.Indeed, the average read rate or a personalbadge was 0.4, and we ound a correlation withcoe cient .37 between a participant’s number o operations and the number o times his or her personal badge was detected. 10 Figure 5bshows how much participants used each tooland application each day relative to that toolor application’s total usage throughout thestudy. We immediately see spikes in roughly the middle o each week (the busiest days) aswell as a gradual decline in usage over the en-tire study. This decline might be in part dueto winter exams in the last two weeks or tothe US Thanksgiving holiday on the last threedays; this might also be the e ect o ading

novelty. The plot also shows that participantsused the Data Browser, Tag Manager, and R d-der more constantly throughout the study, andthe Place Manager and Access Control Inter-

ace more o ten in the beginning o the study,when they were likely to have started de n-ing policies through which riends could viewtheir activities in R dder.

Figure 5c displays the total number o op-erations per ormed through each tool or ap-plication or each participant category. The

columns or Access Control Inter ace andPlace Manager count only metadata updates,insertions, and deletions, omitting the initialselection o existing metadata that occurseach time a user loads these tools. Similarly,Figure 5d shows the operation counts or eachoperation type. The histograms illustrate thatparticipants interacted most requently withtheir data by issuing object and person loca-tion queries via the Tag Manager and R dder.Here, the exit survey revealed that — althoughthe 60 percent o participants without severe

tag-detection problems only rated these appli-

cations an average o 3.2 (standard deviation0.7) on a 5-point scale (where 5 is most use-

ul and 1 isn’t use ul) — most o them (37 o 40) elt that the applications were at least unand novel. Those giving a ranking less than3 o ten cited a lack o tagged objects or par-ticipating riends as a reason or low utility.Most participants also expressed a desire to bepushed the location in ormation, rather thanhave to pull it rom a Web page. The data alsoshows that graduate students de ned 20 o the26 places de ned in the study; this is likely because graduate students occupy a larger region o the building ( ve additional foors)than undergraduates.

The second histogram also reveals someinteresting results about privacy control. Per-haps surprisingly, a participant deleted col-lected RFID data in only one instance. Instead,users ocused on de ning access-control rulesto manage personal privacy. The user who de-leted the data explained in the exit survey thathe did so simply as an experiment to veri y that the tool worked. Furthermore, most users(50 o 67) said that they had very ew privacy concerns involving data collected in our con-trolled exper iment. However, 51 users said they would be more concerned (4 on a 5-point scale,

where 5 is extremely concerned and 1 is un-concerned) i their employer had this data, and56 said they would be extremely concerned i their government had it. Other deletion opera-tions occurred during metadata managementand included minor operations such as delet-ing an object that the system couldn’t e ec-tively track ( or example, a metallic laptop or mobile phone).

Building applications with RFID data in the

IoT is challenging, not just because TREsprovide only low-level in ormation but alsobecause the metadata associated with tags, an-tennas, and events must be personalized andcare ully controlled to create a sa e, meaning-

ul user exper ience. Our Web-based tools aim toempower users by letting them manage meta-data and control privacy.

Based on our study results, we eel thatRFID-based personal object and riend trackingare promising, basic serv ices or the IoT that our tools can quickly enable. One key problem we

must overcome is achieving a su cient den-

Metadata associated with tags, antennas,and events must be personalized andcarefully controlled to create a safe,meaningful user experience.



MAY/JUNE 2009 55


sity o tags and users. Another problem is nd-ing techniques that improve or compensate or low tag-read rates — we’re currently exploringusing stricter tag-mounting strategies as wellas probabilistic data management. 4,10 We alsoconclude that although context-aware accesscontrol seems to be a use ul , easily understoodabstraction or managing location privacy,more evaluation is needed to determine wheth-er it meets user s’ needs when privacy concernsare magni ed. Finally, because 41 out o 67users expressed an interest in using personaltrending applications such as the Digital Diary,we’re studying this application in another, lon-ger user study.

ReferencesComplete RFID Analysis and Forecasts, 2008–2018,1.www.idtechex.com/research/reports/.T. Kriplean et al., “Physica l Access Control or Captured2.RFID Data,” IEEE Pervasive Computing , vol. 6, no. 4,2007, pp. 48-55.

V. Rastogi et al ., “Access Cont rol over Uncer tain Data,”3.Proc. 34th Int’l Conf. Very Large Databases (VLDB 08),

VLDB Endowment , 2008, pp. 821–832.E. Welbourne et. al., “Cascadia: A System or Speci y-4.ing, Detecting, and Managing RFID Events,” Proc. 6th

Int’l Conf. Mobile Systems, Applications and Services

(MobiSys 08), ACM Press , 2008, pp. 281–294.J. Hightower et al., “Learning and Recognizing the5.Places We Go,” Proc. 7th Int’l Conf. Ubiquitous Com-

puting (Ubicomp 05), vol. 3660, Springer, 2005, pp.159–176, 2005.

A. Juels, “RF ID Secur ity and Pr ivacy: A Research Sur-6. vey,” IEEE J. Selected Areas in Comm. , vol. 24, Feb.2006, pp. 381–394.M. Langheinr ich, “A Survey o RFID Privacy Approach-7.es,” Personal and Ubiquitous Computing , Springer, Oct.2008.

A. Nemmaluri et al., “Sherlock: Automatica lly Locat-8.

ing Objects or Humans,” Proc. 6th Int’l Conf. Mobile Systems, Applications and Services (MobiSys 08), ACMPress, 2008, pp. 187–198.M. Philipose et al., “In erring Activities rom Interac-9.tions with Objects,” IEEE Pervasive Computing , vol. 3,no. 4, 2004, pp. 10–17.E. Welbourne et al., “Longitudinal Study o a Build-10.ing-Wide RFID Ecosystem,” to appear in Proc. 7th Int’l

Conf. Mobile Systems, Applications, and Services (Mo-bicomp), June 2009.

Evan Welbourne is a PhD student in the University o

Washington’s Depar tment o Computer Science and En-

gineeri ng. His interests are in sensor networks and datamanagement or mobile and Web-based computing. Heis the lead graduate student on the RFID Ecosystemproject. Contact him at [email protected].

Leilani Battle is pursu ing a BS in computer science and en-gineeri ng at the University o Washington. She workedon the RFID Ecosystem project during summer 2008under the sponsorship o Intel’s Research Experience

or Undergraduates program. Contact her at [email protected].

Garret Cole is a so tware engineer at Microso t Research.His primary interests include sensor networks anddata bases. Cole has a BS in computer science and en-gineering rom the University o Washington. Contact

him at ri [email protected].

Kayla Gould will graduate in June 2009 rom Western Or-egon University with a BS in computer science. Sheworked on the RFID Ecosystem project at the Univer-sity o Washington during summer 2008 under thesponsorship o CRA-W’s Distributed Mentor Program.Contact her at [email protected].

Kyle Rector is a senior in electrical engineering and com-puter science at Oregon State University. She workedon the RFID Ecosystem project at the University o

Washington dur ing summer 2008 under the sponsor-ship o CRA-W’s Distributed Mentor Program. Contacther at [email protected].

Samuel Raymer is pursuing a BS in electrical engineeringat the University o Washington. He is an undergradu-ate researcher on the RFID Ecosystem project and amember o the IEEE Computer Society. Contact him [email protected].

Magdalena Balazinska is an assistant pro essor in the Uni- versity o Washing ton’s Depar tment o Computer Sci-

ence and Engineering. Her interests are broadly indatabases and distributed systems with a ocus on dis-tributed stream processing, sensor and scienti c datamanagement, and cloud computing. Contact her [email protected].

Gaetano Borriello is the Noe Pro essor o computer sci-ence and engineering at the Universit y o Washington.His interests are in the impact technology can haveon solving important social and health problems. Hisrecent work has been in sensing or healthcare andmobile systems or data collection in the developing

world. Contact him at [email protected].

building the internet of things using rfid the rfid ecosystem experience

Documents