By Susan Pynchon, Florida Fair Elections Coalition
November 28, 2005
Fair Elections Coalition's preliminary review of documents obtained in
a Freedom of Information Act request to Florida's Division of Elections
reveals that the state improperly certified the Diebold "paperless" TSX
voting machine and improperly certified Diebold's so-called "blended"
system. Our preliminary findings include the following:
1. Twenty-five percent failure rate The Florida testing of the
TSX took place in March 2005. Four TSXs were supposed to be part
of the testing procedure. However, according to handwritten notes found
in the margins of the testing procedures, "one TSX died." Another note
said that TSX serial #203213 was not used due to a "bad screen."
Further notes indicate that the tests were then conducted on only 3
machines. Although this constitutes a 25% failure rate, no mention of this breakdown (or breakdowns) is made in the final test results.
2. Provisional ballots not private A handwritten note in the margin of the test procedures document says the following: "Note: Review of provisional ballots can occur before ballot acceptance. This needs to change."
This is startling because it indicates that the TSX shows voter
information (name, address, etc.) and how that voter voted before a
decision is made as to whether to accept a provisional ballot. This means that provisional voters do not have a secret vote on the TSX. It would seem that this alone should prevent certification of the TSX. No mention of this problem is made in Florida's official test results report and, in fact, Florida certified the TSX the very next week.
3. Manual Procedures Improper The Reference Guide for the GEMS version 1.18 reveals that manual procedures
are required to define "vote centers" and to accumulate voting results.
These manual procedures mean that the validity of the voting results
ultimately rests on the individual(s) who are implementing these manual
procedures. No amount of testing can cover or guarantee that
these manual processes will be properly implemented. Therefore, the
system is not certifiable because it should not permit manual functions
that cannot be tested and which could affect election results. These
manual procedures are an end run around security features. They may
make the system more flexible and make the supervisor's job easier, but
flexibility is the enemy of security.
Following are the applicable paragraphs from the GEMS 1.18 Reference
Guide, hand-copied from the manual by Susan Pynchon of . We were allowed to view the Diebold manuals but
were told we could not get official copies of them. Based on the
following, no wonder Diebold attempts to keep these manuals private!!
Section 4.5 of the GEMS 1.18 Reference Guide, "Vote Centers:"
"The vote center is the physical location at which ballots are
counted. The results of ballots counted at a vote center are tallied
to the report precinct(s) associated with that vote center. GEMS
automatically creates a vote center of the same name as every polling
report precinct created. However, vote centers must be manually
defined for any cumulative report precincts." (Emphasis added).
Section 4.5.5 of the GEMS 1.18 Reference Guide, "Linking
Report Precincts to Vote Centers:" In order to configure a vote center
with ballot and results tallying
information, the vote center must be linked to report
default, each polling report precinct is created with an equivalent
vote center in the Polling Vote Center category. However, a
jurisdiction may require that vote center-report precinct relationships
be configured differently. Linkages between cumulative report
precincts and vote centers, on the other hand, must be entirely
manually maintained." (Emphasis added).
John Washburn, a computer security expert from Wisconsin, further
explains below why manual procedures, such as those presented above,
should never be allowed:
It is error prone because people make mistakes when dealing
disparate numbers. More problematic it is susceptible to deception. All
accounting frauds have as their essence a fracturing and scattering
of records. This is needed to obscure the misreporting,
miscategorization or embezzlement. These dozens of records at the
precinct level which are supposed to tie out to 1 or 2 county level
numbers is a similar fracturing and scattering of records. Who is
going to tie out and reconcile dozens of numbers from dozens of
precincts late (later than 10:00 pm) on election night? Especially
with an eye toward detecting fraud as opposed to "just getting the
paper work done"?
4. Florida DID NOT TEST the "blending" of the op-scan and TSX
systems Despite several requests, we received no information
regarding the "blending" of the op-scan and TSX systems. In fact,
Florida did not test the blending of these two systems, or if it did,
it is not saying so. In a conversation today, November 28, with a
Division of Elections attorney, Maria Matthews (who has been most
cooperative but who has no way of knowing whether documents are being
withheld), stated, "There is no blended system." When I pointed out
that the system had been certified as the "Diebold 2005B B (Blended) +
(Audio), she said that is "just what they called it" but reiterated,
"there is no blended system." The ramifications of this are huge and
worthy of an entire separate report. In brief:
a. Complicated and Confusing Two side-by-side systems that do not
blend mean two separate sets of poll tapes, two separate uploads from
the precincts (one upload from the op-scan and one upload from the TSX
in each precinct), two separate sets of results, and two complete sets
of computer and audit logs. This is extremely confusing and is a
process wide-open to error and manipulation of election results. If the
systems do blend, why didn't Florida conduct testing on the blending of
those two systems? And why was it certified as a "blended" system?
b. Loss of Private Vote If only one person in a precinct votes on a
TSX, that person's choice of candidates is revealed on the TSX results
tape at the end of the election. If more than one person votes on the
TSX in a precinct, but all those individuals vote for the same
candidates, their votes are revealed on the results tape at the end of
the election. These TSX voters lose their right to a private vote,
because their choice of candidates is revealed. This in fact happened
in the October primary election in Flagler County, Florida, where, in
five precincts, only one person voted on the TSX. In another precinct,
three people voted on the TSX but they all voted for the same
candidate. To repeat, in all these cases, the TSX voters lost their
right to a private vote, because their choice of candidates was
revealed. The right to a secret vote is guaranteed by Section 301 of
the Help America Vote Act, which becomes effective January 1, 2006. A
secret vote is also guaranteed by Florida Statute 101.5606, which
states, "No electronic or electromechanical voting system shall be
approved by the Department of State unless it is so constructed that:
(1) it permits and requires voting in secrecy."
c. Loss of Equal Protection The voters who vote on the op-scan
system have a paper-ballot backup to confirm machine counts of
results. The TSX users vote on a system that has no manual audit
capacity. These voters have lost their right of equal protection as
guaranteed by Amendment XIV of the U.S. Constitution.
d. No user manuals, instructions or security procedures in place for
the blended system There are no Diebold user manuals for the blended
system (we asked). There are no written instructions to elections
officials as to how to accumulate the totals from the two systems; and
there are no security procedures established for the "blending" of the
5. Preferential Treatment of Diebold We are alleging that the Division of Elections has shown preferential
treatment to Diebold and that the Bureau of Voting Systems
Certification created a double standard of certification procedures for
the Diebold TSX and the ES&S AutoMark. As an example of this
preferential treatment, Diebold was required to run 10,010 ballots with
only 138 ballot styles. AutoMark was required to recreate the entire
Miami-Dade county 2004 general election, comprised of almost 800,000
voters, and 3,400 ballot styles. The Bureau of Voting Systems
Certification ran its Phase I and Phase II testing of the Diebold
system from March 21-23, 2005. Six days later the Division of
Elections certified the Diebold TSX and the blended system. It allowed
Diebold to FAX its final application on March 29 and then certified the
system that same day! By contrast, AutoMark successfully completed its
Phase I testing in October, but the Phase II testing is not scheduled
until December 13, 2005. We are alleging that the Division of Elections
is attempting to defeat the AutoMark by delaying its certification at
every step of the certification process, since the Division knows that
all counties must purchase disabled-accessible voting machines no later
than midnight on December 31, 2005.
6. Apparent Cover-Up We believe the Florida Division of Elections,
specifically the Bureau of Voting Systems Certification, is obstructing
our receipt of other information requested and that the Bureau of
Voting Systems Certification appears to be actively involved in a
cover-up of its improper certification of the TSX and the blended
system. We don't know if these problems contributed to the sudden
resignation of Paul Craft, who was until November 21, Chief of the
Bureau of Voting Systems Certification, a department of the Division of
Elections. If this situation did not contribute to Craft's
resignation, it certainly should have.
Florida Fair Elections Coalition is still analyzing other information
received as part of our public information request under the Freedom of
Information Act. We also are aware, and allege, that not all the
information we requested has been provided to us and that it appears
that our efforts to obtain this information is being obstructed by the
Bureau of Voting Systems Certification. For example, we were
previously in possession of two emails from Paul Craft regarding the
"optical scan accumulator adaptor" that were not provided to us. As
another example, Robert Pickett of Diebold has his own Florida state
email address, yet we were provided no copies of emails to or from Mr.
Disclaimer: All information provided above is based on information
provided to and/or our good-faith observations and evaluations.