Generation and Verification of an Optical-Scan Ballot with Audio Assist: This product is made by AutoMARK Technical Systems of Chicago and Lombard, Illinois, and marketed by Election Systems & Software of Omaha, Nebraska. The device provides a blind voter with an audio assist to navigate a touch-screen and select voting choices. (A sighted voter may use the system without the audio assist.) A vision-challenged voter can be assisted also to select a write-in candidate if that is desired. When the voter notifies the voting system that the selection process is complete, an optical-scan ballot is printed. The voter handles the ballot and manually transfers it to the entry slot of an optical-scan sensing unit for summation of the choices with the selections of other voters. The ballot, printed for easy recognition by an automatic sensing unit, may be visually scanned for correctness by a sighted voter, since the entries are human-readable (not encoded). The handling and entering of the ballot into the sensing unit by the sighted voter is sufficient evidence that the voter approves of the ballot as printed. A blind voter cannot read the ballot as he or she handles it, but that type of voter may use an audio assist in the sensing unit to verify that the selections on the ballot are exactly those chosen, making the ballot a voter-approved document for that type of voter. The system has the ability to allow voters with other types of disabilities to select voting choices with a puffing straw or with foot pedals. The system is now being used in Idaho, South Dakota, and Sacramento County, California, according to information on the company's website, www.automarkts.com.
An optical-scan ballot generated by this system is more than equivalent to such a ballot filled out by the voter using a manual method; it has some better features. The system-generated ballot is assured to contain no overvotes nor incorrect markings that would raise questions of "intent of the voter." As with optical-scan ballots filled out manually, these ballots may be recounted by hand or on an independently-managed computer system, and thus provide the basis for an audit, either partial or full.
Copying of the Voter's Final Choices by Capturing the Video Display Signal of a DRE: This device, called VoteGuard, is produced by Democracy Systems of Ormond Beach, Florida (www.democracysystems.com). According to literature from the company, the device receives the video from the "video out" connection of a DRE voting machine. Throughout election day, the VoteGuard device, positioned near the DRE, receives every screen displayed on the DRE as well as all audio output from visually handicapped voters. The company states that "the recording process is an independent, passive recording of the DRE which means VoteGuard in no way communicates directly with the software running on the DRE unit itself." After the close of polls, the company's associated product, called VoteCube Analysis Server, which apparently has computational capability, is called into play. It is used to identify the specific data to be employed in the generation of a set of summary reports. Five types of reports may be produced. In the opinion of this researcher, the necessary requirement for independent verification is the capture of the voter's final set of selections and the independent summation of the choices on them for comparison with the DRE's results. The system does not generate individual ballots. As of the date of this report, the company's released information has not included a notice of any sales to election authorities.
Voter-Verified Audio Audit Transcript Trail (VVAATT): This system was developed by Professor Ted Selker of MIT, Cambridge, Massachusetts. The VVAATT is a voice-operated tape recorder that is connected to a DRE voting machine. While voting, the voter wears headphones. Each time that the voter makes a selection or de-selection on the DRE machine, the system determines the action taken, converts it to audio through voice synthesis, and sends that message to the headphones. The voter is able to verify that the actions that he or she has taken are being recorded correctly. (Norris et al, 2006, pp. 25-26, or search VVAATT on the Internet.)
After the polls are closed, election judges may review the audio tape, identify and sum votes cast for each candidate, and compare the results with those produced by the DRE voting machine.
End-to-End Voter Verification of Election Integrity: This system has been developed by VoteHere of Bellevue, Washington, now a division of Dategrity Corp. It requires a proprietary device called Sentinel associated with each DRE voting machine, as well as VoteHere Management Tools, a computer program mounted on computers dedicated to election auditing at state and local government headquarters. The audit is set up from a central location before the election. The process requires authorities to perform a few tasks on a computer with help from the VoteHere Management Tools. At each polling station, Sentinel devices plug into each DRE, and the Sentinel receives the data about each voter's choices. Pictures of the Sentinel device do not show a screen. According to one description of this system (Norris, Sears, and Nicholas, 2006, pp. 29-34), "In the voting booth, the voter engages in a verification process, producing a receipt that the voter takes home." This verification process involves the use of mathematical cryptography, interaction among the Sentinel, DRE, and the voter, and the printing and displaying of sets of characters generated by the Sentinel. The printing is from the Sentinel and display is on the DRE's screen. The receipt obtained by the voter contains a set of characters that the voter may use on an Internet website after he or she returns home. The receipt is used to verify that the same set of characters is included on a list of identifiers of ballots on the website that have been accepted for tallying. Apparently, the receipt may be received in Braille for the benefit of visually handicapped voters, but this researcher has been told by a representative of the visually handicapped community that only a limited percent of blind persons can use Braille. After the close of polls, the election officials remove a memory stick from each Sentinel and return with it to the dedicated audit computer where the audit of the results is completed.
11. Comparison of Partial versus Complete Independent Verification:
In Section 3 above, the 1969 controversy over the possibility of software fraud was discussed; proposed countermeasures were identified. As a result of this dispute, the state of California decreed that 1% of computer-readable ballots, but in no case less than six precincts, were to be recounted by hand in all election contests. This rule is a requirement for partial independent verification, and it remains in effect at this time. DRE voting systems did not exist in the early 1970s when the rule was put in place.
Voting Systems Using Computer-Readable Paper Ballots: In the report of 1975, this researcher examined the question of whether a 1% manual recount was sufficient to identify malicious software that had the intention of switching votes to change the outcomes of elections (Saltman, 1975, pp. 113-121). It was pointed out that maintenance of control over the use of all ballots printed for the election was a necessary pre-condition for identifying any vote-switching that had been attempted. That requirement remains essential at this time. Vote-switching by a malicious program can be identified only if no other sources of change of results are possible due to mis-allocation of ballots.
The analysis of 1975 to determine a satisfactory percent recount proceeded with the assumption that a malevolent programmer could design a vote-switching scheme that would be difficult to detect by observation by knowledgeable reviewers of past election outcomes in the same jurisdictions. Comparisons of current election results with previous outcomes are often undertaken by political analysts to determine the changes in priorities indicated by changes in voting patterns by voters of the area. For example, knowledgeable observers have used recent past history to counter unsubstantiated claims of fraud in the Ohio presidential election of 2004. A report of the Democratic National Committee, the party of the losing candidate, John Kerry, stated the following:
"That the pattern of voting for Kerry is so similar to the pattern of voting for the Democratic candidate for governor in 2002 is, in the opinion of the team's political science experts, strong evidence against the claim that widespread fraud systematically mis-allocated votes from Kerry to Bush."
If the switching is limited to a very few precincts, and thus must be substantial in each of those precincts, then observers will recognize very unlikely outcomes. Thus, it was noted that:
" ... an alert political party will keep good records of each precinct's voting patterns historically and with respect to similar precincts in the same election, thus minimizing the maximum level of undetectability by observation" (Saltman, 2006, p. 116).
However, clever schemes of vote-switching may attempt to reduce the number of votes switched in each precinct below the "maximum level of undetectability by observation," a parameter used in the 1975 analysis. The mathematical technique used in 1975 was independent sampling without replacement. The model assumed that every precinct contained the same number of voters and that each voter had voted for one of two candidates. The concept was that a malicious programmer had switched results in some precincts but not others, and the question was to determine the percent of precincts to recount such that at least one of the precincts whose votes had been altered would be recounted. The recounting would be expected to expose the fraud. If the switching was spread very thinly over all precincts in order to maximize the likelihood that it would not be noticed, a recount of almost any precinct would identify that a problem existed. It was assumed that, if more than a minimal difference was found between the original reported vote and the recounted vote in any precinct, the supposition would be that the computer program had been manipulated. Then, a decision could be made to examine the computer program, calling all results into question.
Next Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).
