Could electronic voting machine software be programmed to alter the outcome of an election?
Ronald Crane, JD, Software Engineer, et al., in his May 2005 paper titled "A Deeper Look: Rebutting Shamos on e-Voting" available online at Verified Voting Foundation's website, wrote:
"Shamos consistently overestimates the amount of effort required for a dishonest [programmer] to install and operate cheating software...For example, a cheat that moves 1-2% of the votes from one major party's candidate to the other's would be very unlikely to be detected in the vast majority of precincts, as long as it can discern tests from actual voting...
Further, a cheat need not be so unsophisticated as to move a constant fraction of votes. It could observe the voting and move votes only when required, using some sanity checks (e.g., never more than 5% of the votes, never reduce a candidate's total to below some small number, etc.) to avoid triggering suspicions.
...even honest vendors routinely distribute an 'update' to each jurisdiction prior to each election to distribute bug fixes and enhancements. If the voting system uses Shamos's 'ultimate protection against malicious code,' [keeping candidate and party names segregated from the software by presenting them as graphic files] the image-generation program surreptitiously could include the election information."
Aviel Rubin, PhD, Technical Director at the Johns Hopkins University Information Security Institute in their 2003 technical report "Analysis of an Electronic Voting System" in the Johns Hopkins University Information Security Institute Technical Paper TR-2003-19, July 23, 2003, stated:
"We also saw no evidence of any change-control process [the management process for requesting, reviewing, approving, carrying out and controlling changes to a software product] that might restrict a developer's ability to insert arbitrary patches to the code. Absent such processes, a malevolent developer could easily make changes to the code that would create vulnerabilities to be later exploited on Election Day."
Verified Voting Foundation's "Frequently Asked Questions" section of their website (accessed Mar 19, 2009) included the following entry:
"Testing [electronic voting systems] for security problems, especially if they were intentionally introduced and concealed, is basically impossible. Consider the cute surprises inserted by programmers into commercial software that are triggered by obscure combinations of commands and keystrokes, called 'Easter eggs.' These routinely slip through vendor's quality assurance testing, including the amazing flight simulator that is hidden in Microsoft Excel '97. An Easter egg slipped into a voting program would never be detected. If the Easter egg allowed a voter to modify the votes inside the machine, it could change the whole election."
David Jefferson, PhD, Chair of the California Secretary of State's Voting Systems Technical Assessment and Advisory Board posted his July 31, 2005 "Response to the Election Center's Document: DREs and the Election Process" on the website of Georgians for Verified Voting:
"DRE software, and the associated development and distribution processes, should be designed so that attacks are virtually impossible; but sadly, they are not...
Almost certainly 1% or 2%, or in some cases more, of all votes cast nationally on a particular vendor's equipment could be switched without detection, changing the outcome of many races across the country, including even the presidency in a close election such as we had in 1960 or 2000. There seems to unwarranted presupposition that the goal of an attack would be to rig the outcome of some particular election. A much more powerful attack, and one more suited to the vulnerabilities in DRE technology, would be to change the outcome of a large fraction of all very close elections in the nation, without targeting any particular one...
Contrary to intuition of people not skilled in security analysis, making the software ignore the pre-election test or tests and only initiate itself on election day is easy to do...The only kind of pre- or post-election test that would have any hope of detecting a well-designed fraud embedded in DRE code would be a test that looked exactly like a real election to the software, i.e. a test that takes about 13 hours and has the right kind of vote timing and frequency of human errors, the right range and distribution of votes, and had all the other details exactly right. Even with such a carefully-designed test, fraud still might not be detected because cheating could be randomized so that it is not repeatable, and is therefore difficult to distinguish from a timing bug or operator error...
The easiest mode of attack is to tamper with the software while it is in development, or during upgrades, bug fixes, or other software modifications, so that the tainted code is loaded onto the chips through the regular distribution channels. With that kind of attack, one never has to go near the voting machines, or ever set foot in the counties that use them, in order to swing elections."
"GEMS is the central software component of the voting systems developed and marketed by Diebold Election Systems, Inc. (renamed Premier Election Solutions, Inc. in August 2007)... is designed to permit the operator to delete the audit trail records in two important audit logs, intentionally or inadvertently...
Five years after the company recognized the need to remove the 'Clear' buttons from the GEMS audit log screens, not only Humboldt, San Luis Obispo and Santa Barbara Counties in California but jurisdictions in other parts of the country, including several counties in Texas and Florida, continue to use GEMS version 1.18.19.
Regardless of whether Diebold knew that permitting deletion of audit records would violate the federal standards, it clearly knew as early as 2001 that '[a]dding a 'clear' button is easy, but there are too many reasons why doing that is a bad idea.' Deletion of records of an election tally... is not merely a hypothetical possibility... [a Humboldt] county official informed the Secretary of State’s office that, while attempting to print a copy of the Poster Log records from the November 4, 2008, General Election, she inadvertently deleted them instead... the 'Print' and 'Clear' command buttons appear near each other in the same horizontal row."
Michael Shamos, PhD, JD, Distinguished Career Professor of Computer Science at Carnegie Mellon University wrote in his Apr. 2004 paper for the Proceedings of the 14th ACM Conference on Computers titled "Paper v. Electronic Voting Records - An Assessment,":
"Insertion of malicious code by the machine manufacturer has two subcases. In the first, the manufacturer delivers software to a jurisdiction with prior knowledge of the ballot layout, candidate names, etc. for each precinct in the jurisdiction. The machine is programmed to behave perfectly before and after the election but to switch votes to favored candidates during the election. Countermeasures [for such a scenario] are parallel testing and keeping candidate and party names segregated from the software by presenting them as graphic files.
In the second subcase, the manufacturer has no foreknowledge of the details of any specific election but distributes master software that causes candidates of a particular party to win all the future elections. The practical possibility of such a scheme is nil...It is not possible to move a constant fraction of votes from one party to another in each jurisdiction without it being obvious that manipulation is going on...
This nightmare scenario, in which a small number of programmers manipulate the politics of the United States by injecting malicious software into voting machines has more in common with spy novels than it does with reality."
Diebold Election Systems, Inc. released a paper titled "Checks and Balances in Elections Equipment and Procedures Prevent Alleged Fraud Scenarios" on July 30, 2003 on its website as a direct response to the Aviel Rubin et al. paper "Analysis of an Electronic Voting System." The Diebold article stated:
"It would be realistically impossible for a malevolent software developer within the company to successfully insert suspect code. Diebold Election Systems has multiple procedural checks against just such a possibility. Making such changes would require a conspiracy among nearly all the developers in the company, in addition to members of the quality assurance group...[A]ny malicious code would not survive the code review process conducted by the independent testing authority. Even if missed by the ITA, logic and accuracy tests conducted by the local and state entities would almost certainly expose any malevolent behavior.
Furthermore, once malevolent code was inserted, it would be discoverable in perpetuity. It simply would not be possible to install software on tens of thousands of units undetected."
The Election Technology Council, in its "Frequently Asked Questions" section (accessed Mar. 19, 2009), wrote:
"Before the fact vote rigging [is] difficult if not impossible...The concern that unscrupulous programmers will try to rig elections through deceptive software has led to specific processes and policies to avoid such an event. For example, software code passes through numerous internal and external checks before use in an actual election, including rigorous certification testing by independent certification bodies. Voting system software is engineered months in advance of actual elections, making it very unlikely for programmers to know who candidates will be and impossible to know how their names will appear on ballots. The source code is held in escrow by various state and federal officials, and local officials do not have access to it, thus preventing code changes at the local level."
Neil McClure, Vice President of electronic voting machine manufacturer Hart InterCivic, Inc., in his testimony before the US Election Assistance Commission on May 5, 2004, stated:
"[To] alter cast vote records as they are recorded for some desired outcome, malicious software code specific to a particular DRE must be written that shows the voter one outcome that records another...In order for the attacker to remain anonymous and/or perpetrate the attack, the malicious code must remain undetected or hidden.
The malicious code would need to overcome...thousands of different ballot styles in the electronic ballot definition, not including the differences in presentation created by languages or ballot rotation. While the presidential race appears as the first contest, the order of the candidates and the language used can change for different presentations of the ballot, requiring the malicious code to possess control over these functions.
A common claim is that the malicious code need only search for party name and make decisions on vote alteration based on these criteria. In the case of the eSlate [DRE manufactured by Hart InterCivic, Inc.], when instructed by ballot style to display the party name, the party name is converted to a graphic image...The point here is that the selection displayed to the voter has no reference to the party affiliation in the software code...[T]he malicious code becomes much more complex, thereby increasing its size. As the size of the [malicious] code increases, the ability to hide it decreases."