elections and bio metrics

8/6/2019 Elections and Bio Metrics

1/20

Using B iom etricIdentificationTechnologies inthe E lectionProcess


2/20


3/20

Using BiometricIdentificationTechnologies inthe ElectionProcess

Author:Dr. Jim WaymanDirectorU.S. National Biometric CenterCollege of EngineeringSan Jose State UniversityManaged and Edited by:W illiam C. K irnberlingOffice of Election AdministrationFederal Election CommissionPublished by:Office of Election AdministrationFederal Election CommissionWashington, D.C. 20463

September 2001


4/20

Introduction by theOffice of ElectionAdministration

This report is another in the series on In-novations in Election Administration beingpublished by the FEC's Offfice of Election Ad-ministration.

The purpose of this series is to acquaintState and local election officials with inno-vative election procedures and technologies.

Our reports on these innovations do notnecessarily constitute an endorsement by theFederal Election Commission either of anyspecific procedures described or of any ven-dors or suppliers that might be identifiedwithin the report. Moreover, the views andopinions expressed in these reports are thoseof the authors and are not necessarily sharedby the Federal Election Commission or anydivision thereof.

We welcome your comments on these re-ports as well as any suggestions you mayhave for additional topics. You may mailthese to:

Office of Election AdministrationFederal Election Commission999 E. Street, N.W.Washington, D.C. 20463

or else contact usToll free on 800.424.9530 (option#4)Direct on02.694.1095or else by e-mail [email protected].

i i


5/20

Biometric technologies, allowing the auto-matic identification of people using voice pat-terns, eye scans, handwriting style, faces,hands or fingerprints, have been suggested foruse in the election process for eliminatingfraud. Fingerprinting, hand shape, and eyescanning have been used in the United Statesin driver licensing and social service programs.Fingerprinting systems are being introducedinto the election process in several countries,such as the Philippines, Jamaica, Argentina,and Cambodia. W hat are the prospects forintroducing these technologies into our votingsystems?

We will look a t the possible voting applica-tions in this paper and conclude that biomet-ric technologies could be effectively used, evenon a voluntary basis, to detect and deter vot-ing fraud. However, this use would requirefundamental changes in the way we registervoters and would necessitate the creation ofgovernment-controlled databases of physicaland behavioral characteristics of at least somevoters. Although such databases are inherently"fuzzy" and far Less threatening to personalprivacy than phone books or driver's licenses,changes in voter registration procedures toenable biometric data collection could be seenas contrary to the intent of the National Vot-ers Rights Act of 1993 and would likely requireenabling federal legislation.

Using BiometricIdentificationTechnologies in theElection Process

What is Biometric Identification?Biometric technologies use physical charac-

teristics, such as voice tone or hand shape, toidentify people automatically. Behaviors, suchas handwriting style, can also be used by com-puters in this way. The term "identify" is usedhere quite loosely. There is actually nothingin your voice, hand shape or any biometricmeasure to tell the computer your name, ageor citizenship, or to establish your eligibilityto vote. External documents (passport, birthcertificate, naturalization papers) or your goodword establishing these facts must be suppliedat the time you initially present yourself to thebiometric system for "enrollment". At this ini-tial session, your biometric characteristic, suchas an eye scan, is recorded and linked to thisexternally-supplied personal information. Atfuture sessions, the computer links you to thepreviously supplied information using thesame physical characteristic. Even if the bio-metric system works perfectly, the personaldata in the computer, such as your voting eli-gibility, is only as reliable as the original"source" documentation supplied.

Once the computer knows your claimedidentity, it can usually recognize you wheneveryou present the required biometric character-istic. No biometric identification system, how-

1


6/20

ever, works perfectly. Problems are generallycaused by changes in the physical character-istic. Even fingerprints change as cuts, cracksand dryness in the skin come and go. It is farmore likely that the computer will not recog-nize your enrollment characteristic than linkyou to the characteristic of someone else, butboth types of errors do occur.Id ntification: Positive and Negative

Biometric systems are of two types: "verifi-cation" and "identification". We prefer the touse the descriptions "positive identification"and "negative identification" to emphasize theopposite nature of two approaches.

A positive identification system requires youto identify yourself when submitting a biomet-ric measure. Your submitted measure is thenchecked against the measure given when youenrolled in the system to affirm that theymatch. Biometric measures are always "fuzzy"to some extent, changing over time and circum-stance of collection. If the submitted and storedbiometric measures are "close enough", it isassumed that you are indeed the person en-rolled under the identity you claimed. If thepresented and enrolled characteristics are not"close enough", you will generally be allowedto try again. If multiple attempts are allowed,the number of users "falsely rejected" can beunder 1%, although there are always somepeople chronically unable to use any systemwho must be given alternate means of identi-fication. The possibility that an impostor willbe judged "close enough", even given multipleattempts, is usually less than one in ten. Thethreat of being caught in 9 out of 10 attemptsis enough to deter most impostors, particularlyif penalties for fraud are involved.

Positive identification using biometrics canbe made totally voluntary. People not wishing

to use the system can instead supply the sourcedocuments to human examiners each time theyaccess the system.

2


7/20

TABLE 1: IDENTIFICATION: "POSITIVE" AND "NEGATIVE",

POSITIVE NEGATIVETo prove I am someone knownto the system

To prove I am not someone knownto the system

Comparison of submitted sample tosingle claimed template

Comparison of submitted sampleto all enrolled templates

Alternative identificationmethods exist

No alternative methods exist

Can be voluntary Must be mandatory for allBiometric measures linked topersonal information (name, age,citizenship) only through externalsource documents.

Linkage to personal informationnot required.

In "negative identification" applications,found in driver licensing and social service eli-gibility systems where multiple enrollmentsare illegal, a user claims not to be previouslyenrolled. In fact, a negative identification bio-metric system does not require any identityclaim by the users. If a user offers an identity,it is only for the purpose of linking to outsiderecords to establish proof of age or citizenship.The biometric measures themselves cannotestablish name, age, or citizenship and there-fore do not prevent their misrepresentationduring enrollment. These systems do, however,prevent a person from enrolling more thanonce under any identity. Apart from the"honor" system, where each person's word isaccepted, there are no alternatives to biomet-rics for negative identification.

During enrollment, the system must com-pare the presented characteristic to all char-acteristics in the database to verify that nomatch exists. Because of the ongoing changes

in everyone's body, errors can occur in the di-rection of failing to recognize an existing en-rollment, perhaps at a rate of a few percent.But again, only the most determined fraudster,unconcerned about penalties, would take on asystem weighted against him/her with theseodds. False matches of a submitted biometricmeasure to one connected to another personin the database are extremely rare and canalways be resolved by the people operating thesystem.

Negative identification applications cannotbe made voluntary. Each person wishing toestablish an identity in the system mustpresent the required biometric measure. If thiswere not so, fraudsters could establish mul-tiple enrollments simply by declining to usethe biometric system. On the other hand,negative identification can be accomplishedperfectly well without linkage to any externalinformation, such as name or age. This infor-mation is not directly necessary to prove you

3


8/20

are not already known to the system, althoughit may be helpful if identification errors occur.What are the Technologies?

Many biometric methods have been used inpublic systems for "positive identification":hand and finger geometry, iris and retinal scan-ning, voice and face recognition, signature andfingerprinting.

Voice systems have the longest history, dat-ing back to the 1950's. The National Instituteof Standards and Technology (NIST) runs anannual competition for voice recognition tech-nologies using speech sample collected over thetelephone. Error rates (both not recognizing aperson and falsely recognizing a person assomeone else) are a few percent. Error ratesincrease if users do not consistently use thesame telephone. Data storage is a few kilo-bytes of information for each user. Voice rec-ognition is currently being used by the Immi-gration and Naturalization Service at someborder crossings. These systems require nospecial hardware other than a microphone anda computer, so can be quite inexpensive.

Fingerprinting systems have been in use foralmost three decades. Users are required toplace a single finger on a glass "platen" to beelectronically "imaged" (photographed) fromunderneath. Error rates vary considerablybetween vendors and a small percentage ofpeople are unable to use these systems at allbecause of unsuitable fingerprints. The strongadvantage of this method, however, is thatpeople have multiple fingers, each with a dif-ferent fingerprint. By requiring the use ofmultiple fingerprints, error rates can be madequite low for those able to use the system. Datastorage is several hundred to a thousand bytesper user. Fingerprint scanners which link toa computer are now available for under $100

and computer keyboards with built-in scannersare also available. Access control systems us-ing fingerprinting can be seen in prison andmilitary applications.

Hand geometry systems have been in usefor twenty-five years and have seen the larg-est number of fielded applications. The usersplace their right hand on a reflective surfaceand an electronic image is captured of itsshape. No details, such as fingerprints, areseen; only the shape (similar to a shadow) ofthe non-reflecting hand. Error rates for regu-lar users can be considerably under one per-cent, although error rates for infrequent usersare higher due to their unfamiliarity withproper hand placement. A small percentageof people (perhaps 1%) cannot use these sys-tems effectively due to hand irregularities.Data storage is only nine bytes per user. Stand-alone units are available for considerably un-der $1000. The Immigration and Naturaliza-tion System uses hand geometry systems inairports and at border crossings.

Finger geometry is a new method that hasseen extensive use over the last three years atWalt Disney World, where over 10 million sea-son pass holders have been verified. Usersplace fore and middle fingers (like a "peace"sign) into the device, where the general fingershapes (but not fingerprint details) are cap-tured. Error rates for this device are unknown,but its successful use by Walt Disney Worldseems to indicate that a wide range of peoplecan use it effectively. Data storage is 14 bytesper user. The device is available only as "OEM"hardware, thus requiring significant systemintegration.

Iris scanning is a new technology with theadvantage that no direct contact between theuser and the hardware is required. A black-and-white image of the user's eye is recordedfrom a distance. Successful use of this tech-

4


9/20

nology requires trained operators, cooperativeusers and well-adjusted equipment. Experi-mentally measured error rates for this tech-nology are not known, but increase in the pres-ence of uncontrolled lighting conditions. A fewpercent of people seem not to be able to usethis technology. Data storage is about two ki-lobytes per user. Units with a supporting com-puter can be purchased for under $10,000.Several pilot projects using this technology arecurrently underway.

Face recognition is another emerging tech-nology also studied by NIST. For this technol-ogy as well, no direct contact is required be-tween the user and the hardware. Error rateshave been measured at several percent andincrease with time after enrollment to severaltens of percent. Lighting conditions must becontrolled for successful usage. Some percent-age of people are unable or unwilling (such asveiled women) to use this technology. Depend-ing upon the vendor, data storage can be aslow as 256 bytes per user. Software packagesfor under $100 are available for computerswith digital cameras.

Retinal scanning systems have been in usefor fifteen years. Users are required to placetheir foreheads against a head rest and lookat a small light. Error rates, as measured in a1990 test funded by the Federal Highway Ad-ministration, showed no incorrect matches ofdifferent users, but 10% or higher failures tomatch. The percentage of people unable to usethis equipment is unknown. Data storage is72 bytes per user. A few retinal scanners areused inside the Pentagon to control access toclassified spaces.

Automatic signature recognition is of twotypes: static and dynamic. Static signaturesystems attempt to recognize a written signa-ture that has been optically scanned into acomputer. Dynamic signature systems recog-

nize hand movements during the signatureprocess, not the signature itself, and requirespecialized hardware consisting of either a penwith embedded pressure sensor and acceler-ometers or a digital computer pad. Error ratesfor dynamic signature verification were mea-sured in 1992 by a national laboratory. The rateof not recognizing known users was 2% whenmultiple attempts were allowed. Impostorswere accepted at a rate of about 1% under theseconditions. The percentage of persons unableto use this equipment is unknown. Data stor-age requirements range from 50 bytes to 2kbytes per user. Units usable with personalcomputers can be purchased for a few hundreddollars or less.

Although we have wide experience with bio-metric devices in positive identification appli-cations, only two types of biometric methodshave ever been used in documented negativeidentification applications. Those methods arefingerprinting and retinal scanning. Pilotprojects for use of iris scanning for negativeidentification are being planned and/or imple-mented.

5


10/20

Hand geometryFinger geometryVoice recognitionIris scanningRetinal scanningFacial imagingFingerprinting

FingerprintingRetinal scanning

TABLE 2: TECHNOLOGIES DEMONSTRATED IN DOCUMENTEDPUBLIC SYSTEMS

POSITIVE IDENTIFICATIONEGATIVE IDENTIFICATIONWill Biometric Identification Com-promise Privacy?

Whenever biometric identification is dis-cussed, people always want to know about theimplications for personal privacy. If I use abiometric system, will the government, or someother group, be able to get personal informa-tion about me? Biometric measures them-selves contain no personal information. Myhand shape, fingerprints or eye scans do notreveal my name, age, race, gender, health orimmigration status. Although voice patternscan give a good estimation of gender, no otherbiometric identification technology currentlyused reveals anything about me as a person.More common identification methods, such asa driver's license, reveal my name, address, age,gender, vision impairment, height and evenweight! Unlike driver's licenses, however, bio-metric measures cannot be stolen or counter-feited.

The real fear is that my biometric measureswill link me to my personal data, or allow mymovements to be tracked. After all, credit cardand phone records can be used in court to es-

tablish a person's activities and movements.There are several important points to be madeon this issue.

Only biometric measurements which I havesurrendered to a system through "enrollment"will be known to that system. If I have neverenrolled (given my fingerprint with support-ing identification documentation) in a finger-print system any use I make of a fingerprintsystem cannot be linked to 'me' (my identity).

Biometric measures cannot generally betaken without my knowledge, so I cannot beenrolled in any system without my participa-tion. Exceptions are face and voice patterns,which can be taken without my knowledge."Latent" fingerprints left on surfaces can be"lifted" by those trained in investigative tech-niques, but such prints are generally not of aquality suitable for enrollment purposes inelectronic systems.

Phone books are public databases linkingme to my phone number. These databases areeven accessible on the Internet. "Reverse"phone books also exist allowing my name tobe determined from my phone number. Even

6


11/20

if I have an unlisted number, my number andall information on calls made from that num-ber may be available to law enforcement agen-cies through the subpoena process. There areno public databases, however, containing bio-metric identifiers, and there are only a few lim-ited-access government databases containingbiometric measures. Eight States have elec-tronic fingerprint records of social service re-cipients (AZ, CA, CT, IL, MA, NJ, NY, TX ), fiveStates (CA, CO, GA, HI, TX) maintain elec-tronic fingerprints of all licensed drivers',nearly all States maintain copies of driver'slicense and social service recipient photos, theFBI and State governments maintain finger-print databases on convicted felons and sexoffenders, and the federal government main-tains hand geometry records on those who havevoluntarily requested border crossing cards.General access to this data is limited to theagencies that collected it, but like credit cardand phone "toll records", this information canbe released or searched by law enforcementgroups acting under court order.

Unlike your Legal name and your Social Se-curity, credit card and phone numbers, yourbiometric measures are rather fuzzy and in-exact, being somewhat different every timethey are measured. Further, your biometricmeasures will be rather similar to the biomet-ric measures of others. Consequently, even ifyou could gain access to a database contain-ing biometric measures, they could not be "re-versed" like a phone book to reveal names fromidentifying numbers. Two technologies, elec-tronic fingerprinting and retinal scanning,have been objectively demonstrated to be ex-ceptions to this reversibility rule, if data iscarefully collected from cooperating users. So

if you want to discover someone's identity, thebest way is with a phone number, not a bio-metric identifier. If you want personal infor-mation about someone, start with their name;a biometric identifier will be of no help.

Biometric identifiers in databases of driv-ers, social service recipients or border cross-ers, are far less distinctive than the names,addresses and ID numbers also in these data-bases, and do not allow users to be tracked ormonitored like credit card and phone numbersdo. For this reason, databases of drivers andsocial service recipients are always indexed byname or identification number even if theycontain a biometric record. Biometric identifi-ers are nearly impossible to steal or falsify thanthese other identifiers, allowing protectionfrom identity theft or impersonation. In con-clusion, adding a biometric identifier to cur-rent voter registration databases would notpresent any privacy risk to any voter, but couldbe used to prevent or deter privacy loss throughidentity theft.

1 WV maintains a voluntary fingerprint database on drivers who wish to use biometric identification.

7


12/20

TABLE 3: BIOMETRICS AND PRIVACY

1) Unlike more common forms of identification, my biometric measures contain nopersonal information about me and cannot be stolen or forged.2 ) Some biometric measures (face images, voice signals, and "latent" fingerprints left

on surfaces) can be taken without my knowledge, but can't be linked to me without apre-existing database,

3) The federal government maintains a fingerprint database on convicted felons andsome State governments maintain fingerprint and image databases on drivers andsocial service recipients.

4) My social security or credit card number, and sometimes even my legal name, canidentify me out of the entire U.S. population. This capability has not beendemonstrated using any single biometric measure.

5) Like phone and credit card information, biometric databases can be searched outsideof their intended purpose by court order.

6 ) Unlike your credit card, phone or Social Security numbers, biometric characteristicschange from one measurement to the next.

7) Searching for personal data based on biometric measures is not as reliable or efficientas using identifiers like your legal name or your Social Security number.," IM4NNNVAikeWv.. .1.3, , ,, , , ,, , ,o34,y, , , ,. . , .. . , ,, , , ,, , , ,, , . ., , , ,, , ,N1-aaa, ,IMMIMWMOSOM MISI,KWA,, , ,VRANAW.SMNIRO101241M101,16MSSOM OOMMOSAM S1tiegiMRANNWHARVAVNEMAINONNW MIRARM,OloWV.

Election System GoalsBiometrics have been successfully used to

increase the integrity of the driver's licensingand social service benefit distribution pro-cesses in many States. There is no questionthat it is technically possible to use biomet-rics to limit fraud in voting processes as well.The 14th, 15th, 19th , 24th, and 26th Amend-ments to the U.S. Constitution establish vot-ing as the right of all citizens 18 years of ageor older who have not been convicted of a dis-qualifying crime. The recognition of voting asa "right", however, separates it from the iden-tified "privileges" of driving and receiving so-cial service benefits

Further, by federal law we have adopted thepotentially competing goals of limiting fraud

and enhancing voter registration. The Na-tional Voter Registration Act (NVRA) of 1993seeks: (1) "...to increase the number of eligiblecitizens who register to vote in elections forFederal office....; (2) (to enhance) the partici-pation of eligible citizens as voters in electionsfor Federal office; (3) to protect the integrityof the electoral process; and (4) to ensure thataccurate and current voter registration rollsare maintained."

It is fair to say that these are the goals ofthe current Congress as well and should beour goals when suggesting changes to the vot-ing process. Protecting the integrity of the elec-toral process should include making sure thatonly eligible voters register, and that only reg-istered voters cast vote. It seems clear that

8


13/20

personal identification, possibly involving bio-metrics, is a key element here. The challengewill be to protect the integrity of the processwithout burdening this right to vote in waysthat may decrease registration by eligible vot-ers.Making Sure Only Eligible VotersRegister

The 14th, 15th, 19th , 24th, and 26th Amend-ments to the U.S. Constitution identify eligiblevoters as all citizens 18 years of age and olderwho have not been convicted of a disqualify-ing crime. Implicit in these Amendments andthe NVRA, and explicit in voting codes, is theadditional requirement that each citizen is eli-gible to register only once. Establishing thatyou are a citizen at least 18 years old cannotbe done directly by biometric identification.This requires trusted source documents, likea certified birth certificate or a passport. Ifthese source documents were linked to a bio-metric record, which they are not, positive bio-metric identification could be used to estab-lish the connection of the presenter to the pre-sented source documents. Driver's licenses inthe States of Texas and Georgia display en-coded fingerprints and could be used to link apresenter to a n identity through biometrics,but they are not proof of eligibility to vote andmerely transfer the original identification bur-den to the driver's licensing system. In theabsence of biometric data on passports or birthcertificates, biometric identification cannot beused to establish my eligibility to vote.

Each citizen is allowed to register only onceand in one jurisdiction: "one voter, one vote".In registering to vote, I declare my previous

registration, if any, so that I can be removedfrom the voter roles of my previous jurisdic-tion. Negative biometric identification couldbe used to determine if I am previously regis-tered in the current or other jurisdictions, pre-venting voter fraud through multiple registra-tion of the same voter.

Under the 14th Amendment, citizens canlose eligibility to vote for conviction of somecrimes. In registering to vote, I attest that Iam not someone who has lost eligibilitythrough conviction of a disqualifying crime.The National Association of Secretaries ofState has recommended that a task force in-vestigate the creation of a national clearing-house of names of disqualified voters to allowthe cross jurisdictional enforcement. Thisnegative biometric identification could be donewith fingerprinting because fingerprint recordsare available on those convicted of disqualify-ing crimes.

In considering biometric identification forpreventing multiple registrations or for pre-venting registration of disqualified voters, re-call that such "negative" identification mustalways be mandatory for all enrolling in thesystem. In other words, enforcement of "one-voter, one vote" and disqualification provisionsusing biometrics would require the mandatorybiometric measurement of all registration ap-plicants. In the case of preventing registrationof those disqualified by criminal record, fin-gerprinting of all registering voters would berequired. This would not only require special-ized equipment', it would place a burden onthe entire process for all registrants to deterthe activities of a very few. Further, manda-tory fingerprinting might be considered a de-

'There are "ink-less" fingerprinting systems available, but there is no evidence that such systems can besuccessfully used, except by forensic experts, to acquire fingerprints suitable for electronic systems.

9


14/20

terrent to registration by those who mistak-enly believe that fingerprint databases on mi-nor traffic offenders exist through driver's li-censing systems.

Burdening the entire process should be con-sidered only if there is adequate documenta-tion of a clear need. We know of no documentedstudies on a national basis showing massivefraud through multiple registrations orthrough the registration of criminally disquali-fied voters. In short, it is not clear that thereis a currently justification for adding manda-tory "negative" biometric identification to thevoter registration process.Making Sure Only Registered VotersVote

Another identification problem faced in thevoting process is the positive identification ofvoters at the polls. Can the poll workers becertain that people appearing at the polls arewho they claim to be? The current solution tothis problem in many jurisdictions is to havevoters announce their name aloud; the conceptbeing that poll workers or other voters presentcould challenge false claims of identity. Vot-ers are also required to sign a roster. If a voter'sidentity is challenged later, the roster signa-ture can be compared to that given at regis-tration.

This process could be strengthened in anumber of ways. Voters could be asked at thepolls to supply additional information given atthe time of registration, such as their addressor birth date. Voters could be asked to presentidentification documents, such as a driver's li-cense, birth certificate or utility bill. Voterscould be asked to bring to the polls mailed elec-tion materials showing name and address. Or

voters could supply biometric identification.This use of biometrics for positive identifi-cation could be done on a voluntary basis. Ju-

risdictions wishing to give voters this optionwould allow those requesting biometric iden-tification to record a biometric measure whenthey register. This would require special equip-ment at the registration sites, as well as at thepolling places. It would require the centralizedstorage of these measures by the jurisdiction.It would also require the transmission of thebiometric measures between the jurisdictionand the polling places on election day.Of all the methods we've listed here forstrengthening the process of identifying vot-ers at the polls, biometric identification wouldrequire the most additional equipment andcause the most changes to the current systems.However, it would also be the method hardestto defraud. We have, again, seen no docu-mented evidence showing widespread, nationalproblems with voter identification at the polls.If there is a need to strengthen the system ina particular jurisdiction, it seems sensible tostart with other less secure and less costlymethods of voter identification. Only afterthese methods prove to be insufficient, or thereis a general demand by the voters to allow sub-stitution of biometrics for these methods, coulda practical case be made for biometric identi-fication.Absentee, Nomination and PetitionApplications

The identities of voters applying for absen-tee ballots, petitioning the government ornominating candidates are currently verifiedby comparing the signatures on these docu-ments to signatures in the voter registration

10


15/20

rolls. This labor-intensive process is oftenaided by electronic "election signature re-trieval" systems. Handwritten signatures fromvoter registration documents are opticallyscanned into a computer system. Then, elec-tion officials can electronically recall these sig-natures to compare them to those on petitions,absentee ballots and nomination forms. Theactual comparison of the signatures is done byhuman eye.

This process of comparing signatures couldbe automated. Computer programs for com-paring written signatures currently exist inlaboratories, but are not currently commer-cially available. These systems require no spe-cial hardware and are different from commer-cially marketed "dynamic signature verifica-tion" that require special pens and tablets.Even if quite crude, this form of biometric iden-tification could successfully reduce the humanworkload by automatically accepting the sig-natures that are clearly legitimate or at leastvery good forgeries: the same signatures thatwould be easily accepted by human examin-ers. Only signatures that are not obviousmatches would require a human decision. Webelieve that such automated signature match-ing could be profitably integrated into currentelectronic signature retrieval systems.Other Applications

We can imagine more elaborate uses of bio-metrics for prevention of "chain balloting" orto allow completely anonymous voting. Chainballoting is a method for corrupting document-ballot elections. A campaign worker gives thecomplicit voter a pre-marked ballot before he/she enters the polls. At the polls, the voterconceals the pre-marked ballot and is given ablank ballot. The pre-marked ballot is cast and

the blank ballot surreptitiously returned to thecampaign worker after leaving the polls. Thecampaign worker marks the ballot for the nextvoter. In 1992, about half of the States using adocument ballot had procedures in place toprevent chain balloting.

Biometric identification could be used toprint a biometric identifier on the ballot stubwhen the ballot is issued. The biometric mea-sure on the stub could be compared to onetaken from the voter when the vote is cast. Thestub would be given to the voter so that no bio-metric record of the voter would remain at thepolls after the voter has left. This applicationwould require the mandatory biometric mea-surement of all voters.

In theory, completely anonymous votingcould be accomplished by registering volun-teering voters under a biometric identifier.Eligibility at registration would be ascertainedusing current methods and registration recordswould include the voter's name. Only thevoter's biometric identifier, however, would besent to the polls. At the polls, voters wouldpresent the biometric identifier in lieu of an-nouncing a name. This extreme applicationwould significantly alter the current systemof publicly releasing the names of those whohave voted.Internet Applications

In 1999, the State of California created an"Internet Voting Task Force" to study the pos-sibility of casting votes over the Internet. Thetask force found that one of the obstacles toInternet voting would be the identification ofthe person casting the vote.

11


16/20

The problem of identification of Internetvoters is one of both positive and negative iden-tification. Negative identification would berequired if we wished to prevent multiple reg-istrations of the same person. Positive identi-fication would be required to identify the per-son casting the vote as the registered voter.

As discussed, negative identification mustbe mandatory for all voters. In the case ofInternet voting, multiple Internet registrationscould be prevented by the mandatory biomet-ric identification of all Internet voters at reg-istration. This would not require mandatoryidentification of non-Internet voters if we werewilling to a llow for the possibility of fraudthrough both Internet and paper registrationof the same voter under different identities.Internet registration with the submission of abiometric identifier could not be securely doneover the Internet, but would require "in per-son" registration and the collection of the bio-metric identifier by trained and trusted per-sons This identifier would be placed in a da-tabase under the control of the jurisdiction.Upon verification that the registering voter isnot already in the database, a voter ID num-ber, code or PIN could be issued. Biometricidentification and specialized hardware at thetime of voting would not be required for nega-tive identification.

Positive identification by Internet votersusing biometrics would require that biometricmeasures be previously registered "in person"with the jurisdiction and would require stan-dardized biometric collection hardware andsoftware on the computer used for voting. Posi-tive biometric identification might be used ona voluntary basis to replace other types of PINor password identification. An added problemis the occasional failure of all biometric tech-niques to recognize properly registered users."Provisional" voting would have to be allowed

in cases where the voter's submitted biomet-ric measure did not seem to match the regis-tered measure.

In short, biometric identification couldbe an important adjunct to Internet voting, butwould not solve all identification problems in-herent in Internet voting.Conclusions

In this paper we have looked at specific ap-plications of biometric technologies to the vot-ing process. We can conclude that biometricidentification could be effectively used, evenon a voluntary basis, to detect and deter vot-ing fraud. Biometric identification, however, isnot a "silver bullet" capable of solving all prob-lems of voter identification without any unde-sirable side effects. Use would require funda-mental changes in the way we register votersand would necessitate the creation of govern-ment-controlled databases of physical and be-havioral characteristics of at least some vot-ers. Although such databases pose no threatto the privacy of voters, the process could beseen as an additional burden on the registra-tion process. We would need to carefully con-sider the potential impact of such changes onthe competing requirements of the NationalVoter's Rights Act of 1993 to both enhance voterparticipation and to protect the integrity of theelectoral process.

12


17/20


18/20


19/20

Other titles in our Innovations series include:n Vol 1: The Voting Authority Cardn Vol 2: Optical Scanning Technology for Purposes other than

Ballot Countingn Vol 3: Election Signature Retrieval Systemsn Vol 4: Using NCOA Files for Verifying Voter Registration Listsn Vol 5: Agency Voter Registration Programsn Vol 6: Motor Voter Registration Programsn Vol 7: Mail Registration Programsn Vol 8: Election Document Retention in an Age of High Technologyn Vol 9: Early Votingn Vol 10: Ballot Security and Accountabilityn Vol 11: All-Mail-Ballot Electionsn Vol 12: The Electronic Transmission of Election Materia lsn Vol 13: Simplifying Election Forms and Materialsn Vol 14: Recruiting Poll Workersn Vol 15: Ensuring the Accessibility of the Election Processn Vol 16: Using the Internet in Election Officesn Vol 17: Acquiring Election Systems and Equipment


20/20

For information about otherInnovations in Election Administration

contact

Office of Election AdministrationFederal Election Commission999 E . Street, N.W .W ashington, D.C. 204 63

Toll Free 800.424.9530 (option #4)[email protected] ELECTION COMM ISSIONWASHINGTO N, DC 20463 Bulk Rate MailPostage and Fees PaidFederal Election Com missionPermit Number G-31OFFICIAL BUSINESSPenalty for PrivateUse, $300

elections and bio metrics

Documents