Insight from the mind of Randall Munroe.
Category: electronic voting
Insight from the mind of Randall Munroe.
Computer vote counting is a radically different trust model than a hand-counted election.
Instead of a vote counted in public by known individuals, with observers, you have a third-party for-profit vendor counting the vote in private, with testing by the election authority, but no meaningful observation.
If an elections authority proposed to pay a vendor’s employee to count votes in private, even with a complete background check of the employee, I have the feeling that not many people would go for it.
But in what is essentially the same scenario, except with the employee replaced with a “machine”, people don’t seem to have a problem.
I thought about why this might be the case, and it seems to one primary and one secondary thing. Primary is the idea that a person has unlimited freedom of action, but a “machine” does not. Secondary is the confusion that because the vote tabulator itself is in public, somehow the vote count is still “in public”, even though it’s taking place inside the literal black box of the tabulator.
This is I guess a 20th Century collision with 21st Century realities. If you have an assembly line with a machine that makes pins, if you turn your back, it won’t suddenly decide to secretly make hammers. Because the vote tabulator looks like some sort of machine, and is described usually as either “electronic” or “machine”, people think it is a single-function device. But it’s actually a general purpose computer. Which means that not only does it have a wide range of freedom of action, just like a human being, it can lie to you about what it is doing, just like a human being.
It would be interesting to see a polling station set up with a giant human-sized black box that the ballots go into to be counted, and see how people reacted to that. Because there really is no difference between that and the computer vote tabulator. Basically you’ve taken a very limited trust in known people you can watch in public, and changed it to a very extensive trust in unknown vendor employees and in the elections organisation itself operating in private.
If you have a very complicated count and very high expectations of a fast count, then there is some justification in using a vote counting computer, as long as you don’t trust the computer. You have to audit the paper, not the computer. You can test the computer as much as you want, it can always lie. This is exactly what happened in the Volkswagen diesel emissions scandal, where the car’s computer could detect when it was being tested and would change its behaviour accordingly. So when you use a computer to count paper, you have to audit the paper with a manual count (a risk-limiting audit). Unfortunately as far as I know, no Canadian jurisdiction follows a computer ballot count with a risk-limiting audit.
In any case, Canadian federal and provincial elections are trivial to count. You literally just sort the ballots into a few piles. And because the count is simple it is also fast.
The Ontario provincial switch to vote counting computers is wrapped with PR about technology, but it’s actually about staffing. (The underlying concept is literally called "Proposal for a technology-enabled staffing model for Ontario Provincial Elections".) Basically it’s hard to get people to staff elections now, and they’re tired by the end of the day which means they are sometimes not in the best shape to do a bunch of precise counting. There are many many ways to address elections staffing. For example, you could simply bring in people, e.g. High School students, to do the count at the end of the day.
Addressing a staffing problem by completely changing the counting trust model wouldn’t have been my choice. And I would assert that the only reason it’s even possible is because people don’t realise the trust model has been radically changed.
In any case, online voting is a much much worse problem that vote counting computers, so this is about all I have to say about the vote tabulators issue.
May 11, 2018 2018 Ontario Provincial Election to use vote counting computers
The 2018 Ontario Provincial Election taking place on June 7, 2018 will for the first time use vote counting computers province-wide. This replaces hand-counting of ballots.
The computer vote tabulators use optical scan technology to read hand-marked paper ballots.
This is the least-worst use of computer technology for vote counting as the hand-marked ballots are still available to be counted. However, these are still computers that have to be programmed, which means there is always the potential for errors or malicious code.
Fundamentally in elections, you don’t trust anyone. That means you don’t trust the computer vote tabulator either. Use of computer vote tabulators introduces the following key questions:
The new voting procedures were launched with a May 9, 2018 press release (PDF) and accompanying media event.
Elections Ontario is modernizing the voting process and putting the needs of electors first by introducing technology in the polls. Election officials will be using electronic poll books (e -Poll books) and vote tabulators across the province for advance voting. On election day, 50% of the polls will have vote tabulators and e-Poll books … serving 90% of electors.
There was a Canadian Press story by Liam Casey, see e.g. CBC News – Ontario to use electronic voting machines for first time in spring election – May 9, 2018.
The tabulator is a Dominion Voting ImageCast® Precinct computer optical scan vote tabulator.
The history is buried in the post-event reports for two byelections that tested the technology:
It is very clear from the Proposal that the key issue is staffing; the technology is being introduced to address poll staffing issues.
Disclaimer: I am not a lawyer.
Additional questions raised by the use of computer vote counting equipment:
In the March 22, 2018 Guelph Mercury article Ontario’s voting system secure, chief election official says the following statement is made by the Chief Electoral Officer:
“The Ontario government has hired a cybersecurity team to assist any of the ministries with private security — and we’ve been working with that team over the last year, year and a half, and they’ve been working with all of our systems,” he said.
“They’ve been doing penetration testing, vulnerability testing … to ensure that our systems are up-to-date and secure. There have been some slight alterations based on their recommendations, and we are very confident and we take security very, very seriously.
“I want to make sure that all the systems and all the personal information that we have is protected.”
For more about what it means to change from public hand-counted ballots to ballots counted by a computer from a private for-profit company, see computer vote counting is a radically different trust model.
The governing law is the Ontario Election Act, R.S.O. 1990, c. E.6
The relevant sections, modified in 2016 (Election Statute Law Amendment Act, 2016, S.O. 2016, c. 33 – Bill 45) and in force as of January 1, 2017 are:
Next section blockquoted due to complexity:
Restrictions re equipment
4.5 (3) The following restrictions apply with respect to the use of vote counting equipment:
1. The equipment must not be part of or connected to an electronic network, except that the equipment may be securely connected to a network after the polls close, for the purpose of transmitting information to the Chief Electoral Officer.
2. The equipment must be tested,
i. before the first elector uses the equipment to vote, and
ii. after the last elector uses the equipment to vote.
3. For the purpose of paragraph 2, testing includes, without limitation, logic and accuracy testing.
4. The equipment must not be used in a way that enables the choice of an elector to be made known to an election official or scrutineer.
The only section that speaks about voting equipment security appears to apply solely to section 44.1 Accessible voting equipment
Accessible voting equipment, etc.
44.1 (1) At an election, accessible voting equipment and related vote counting equipment shall be made available in accordance with this section and in accordance with the Chief Electoral Officer’s direction under subsection (2). 2010, c. 7, s. 24 (1).
(5) Despite subsection (1), accessible voting equipment and related vote counting equipment shall not be made available unless an entity that the Chief Electoral Officer considers to be an established independent authority on the subject of voting equipment and vote counting equipment has certified that the equipment meets acceptable security and integrity standards. 2010, c. 7, s. 24 (1).
There is no analogous section under 4.5 vote counting equipment. Disclaimer: I am not a lawyer.
April 30, 2018 – 42nd Parliament, 1st Session – Bill C-76 Elections Modernization Act
The proposed changes to section 18.1:
It’s a bit unclear what the difference is between 18.1(2) “alternative voting process” and 18.1(3) “voting technology”. Can an alternative voting process include new technology? I have to assume so, particularly given how it is framed in the Chief Electoral Officer’s recommendations. (There is no definition provided in the bill for “alternative voting process”).
In An Electoral Framework for the 21st Century: Recommendations from the Chief Electoral Officer of Canada Following the 42nd General Election, Table A—Recommendations Discussed in Chapters 1 and 2, A15. 18.1 it says “The distinction between the approval requirement for testing an electronic voting process and any other alternative voting process should be removed”.
2014, c. 12, s. 8
15 Sections 18.01 and 18.1 of the Act are replaced by the following:
18.01 The Chief Electoral Officer may provide assistance and cooperation in electoral matters to electoral agencies in other countries or to international organizations.
18.1 (1) The Chief Electoral Officer may carry out studies on voting, including studies respecting alternative voting means.
(2) The Chief Electoral Officer may devise and test an alternative voting process for future use in an election.
Voting technology — electors with a disability
(3) The Chief Electoral Officer shall develop, obtain or adapt voting technology for use by electors with a disability, and may test the technology for future use in an election.
(4) Neither an alternative voting process nor voting technology tested under subsection (2) or (3) may be used in an election without the prior approval of the committees of the Senate and of the House of Commons that normally consider electoral matters.
Clause 15: Existing text of sections 18.01 and 18.1:
18.01 The Chief Electoral Officer may, at the Governor in Council’s request, provide assistance and cooperation in electoral matters to electoral agencies in other countries or to international organizations.
18.1 The Chief Electoral Officer may carry out studies on voting, including studies respecting alternative voting processes, and may devise and test an alternative voting process for future use in a general election or a by-election. Such a process may not be used for an official vote without the prior approval of the committees of the Senate and of the House of Commons that normally consider electoral matters or, in the case of an alternative electronic voting process, without the prior approval of the Senate and the House of Commons.
Prince Edward Island (PEI) – 2016 Plebiscite on Democratic Renewal – Voting Integrity Audit Report – from the Independent Technical Panel on Voting Integrity (ITPVI) – November 30, 2016
This report is Section 3 Appendix in the 2016 Annual Report of the Chief Electoral Officer of PEI (PDF), starting on page 35.
Section 11 of the Voting Integrity Audit Report is Considerations for Applying E-Voting Options [Internet voting] in Canadian Public Elections.
The report recommends against Internet voting at the federal and provincial levels, except for absentee voters.
There is a need to maintain an acute level of awareness of the risks to electoral integrity that these new voting methods present. The implications of a breach of the public trust that exists today suggests strongly that internet and telephone voting in Canadian provincial and federal parliamentary elections be considered channels that should be limited to use only by absentee voters for the immediate foreseeable future. …
It is important that leaders in Canadian electoral administration manage public expectations and articulate their concerns about the fact that a perfectly secure and fool-proof electronic voting system does not yet exist.
This recommendation was picked up in the news media, e.g. CBC News PEI – Online voting not ready for federal, provincial election: officials – May 4, 2017.
The group concluded a high-stakes provincial or federal election could attract groups looking to intervene in illicit ways through cyber-attacks, hacking or other means.
The report also does an excellent job of showing the “additional risks and controls associated with online electronic voting” [Internet voting]. These include (highlighting by me):
1. Trusted digital voter identification and authentication is a requisite additional control. An irrefutable digital identity is the first safeguard in ensuring that eligible voters can vote (and can vote only once), and in ensuring that ineligible voters are not permitted to vote. Establishing this identity with a robust ‘shared secret’ is a mandatory prerequisite.
2. The onus is on the buyers, designers, developers, maintainers and operators of any electronic voting system to demonstrate rigor in the specifications, certifications, accreditations, testing and operation of the e-voting system to ensure it is able to mitigate the full range of risks to a reasonable and acceptable level. This has to be achieved to a level of satisfaction regarding both hardware and software risk mitigation. The remaining level of risk needs to be accepted by all stakeholders.
3. With the elimination of the controls that were previously implemented in manually controlled voting processes (refer Appendix ‘G’: Controls C1 – C5), traditional risks are not as fully mitigated as before. In fact, the following risks are difficult to mitigate in any meaningful way:
a. Vote buying / vote secrecy (“I’ll just take a selfie in front of my screen”)
b. Voter coercion (Unless reported, it is impossible to determine if a vote is being coerced)
4. The risk of a voter voting with stolen credentials can only be partially mitigated by effective voters list management and the implementation of a trusted digital voter identification and authentication scheme. Digital voter identification must be robust, but it must also be easily managed so as not to become a barrier to voting because it is overly complex for a voter to use as seldom as once every four years.
5. The additional risks of compromised end-user hardware or software, or a broad regional or national attack on internet infrastructure, remain unmitigated.
The report also identifies the extremely high standard to which we must hold Internet voting, as the transparency provided by conducting paper ballot voting and counting in public are lost when using completely computerized processes. Highlighting added by me.
The onus is also completely on the online electronic voting system implementer to ensure that controls are established within the e-voting system that meet the legislative requirements of the jurisdiction, and provide an adequate level of transparency for all stakeholders. Simply depositing electronic votes into a ‘black-box’ where they are stored and counted is unlikely to meet stakeholder demands for maintaining a high level of public confidence, unlikely to publicly show that voting risks are continuing to be
managed responsibly, and unlikely to prove to candidates and political parties that the electoral process and controls continue to deliver a trusted and accurate result.
SIDEBAR on turnout:
A demonstration of the reality of Internet voting turnout was the 2016 Prince Edward Island Plebiscite on Democratic Renewal which had 10 days of online voting in addition to two days of in-person voting. Not only was the overall turnout low at 36.5%, but the turnout for ages 18-24 was the lowest of any age range, at 25.47%.
Numbers from McLeod, G. B. (2016, November 9). Interim Report of the Chief Electoral Officer for the 2016 Plebiscite on Democratic Renewal. http://www.gov.pe.ca/photos/original/elec_demrefpleb.pdf
The Government of Wales is running a consultation: Electoral reform in local government in Wales. The consultation closes 10 October 2017.
A variety of questions are considered, but for the purposes of this blog there are three of interest:
In what I have found is fairly typical fashion, the main consultation paper (PDF) does not cite any references, and makes brief, broad, generally positive statements. (The youth and “easy read” consultation versions in turn simplify and amplify these statements to an extreme degree.)
You can fill in an online form,
but in order to be able to provide more extensive comments, you may instead want to download the email response form (DOCX), complete it (or complete whichever sections are relevant to you) and send it to RLGProgramme@wales.gsi.gov.uk
Reminder that the deadline is 10 October 2017.
(page 18 in main consultation document)
This is defined solely as touch-screen voting. There is no mention of paper output, so presumably paperless touch-screen voting.
5.14. This implies the installation of equipment at polling stations (and possibly other locations) to enable touch-screen voting. …
5.15. Electronic voting is already used widely internationally, particularly in India but also in Belgium and Estonia amongst others.
I think this is a misunderstanding of voting in Estonia. As far as I know, Estonia doesn’t use paperless touch screens. On voting day, voting is on paper.
There isn’t any serious examination of security risks to voting machines (voting computers), but there is the rather extraordinary assertion that electronic voting could lead to less challenging of “votes” (presumably this means fewer challenges to election results).
5.19. … there would need to be secure procedures in place to ensure the security of data being transmitted from the polling places to the central count operations. The challenging of votes could become less likely. …
I, on the other hand, think paperless touch-screen voting would introduce not only high security risks, but would make challenges to election results both more likely and impossible to satisfactorily resolve (as there is no physical trail to audit).
(page 19 in main consultation document)
It’s clear this means Internet voting.
5.20. This refers to a process of voting through access of the internet by an electronic device, using an individual recognition code. The use of codes of different sorts to ensure that only the intended person is accessing a system is now commonly used for purchasing, banking, voting in elections within political parties, trade unions and other organisations. Registration to vote is now routinely performed online, as is registering/taxing a motor vehicle and accessing a multitude of other public services or transactions.
Where to begin? Voting doesn’t have the same requirements as banking; voting has much harder to satisfy requirements as the transactions have to be anonymous and aren’t reversible. Voting is not a regular online personalised transactional service.
5.21. Remote voting was piloted in local elections at South Buckinghamshire in May 2007. Although only a minority made use of the facility, 10 years later the option is likely to be more popular. There were no particular technical difficulties but the Electoral Commission called for the pilots to be suspended – along with all others – until the system was generally more secure. There is a risk that, with registering being done remotely, fictitious voters could be created and that voting might not take place in secure environments. In addition, realistic concerns exist about cyber security, and any system needs to be as secure as possible from the dangers of hacking and manipulating votes. This must be weighed against this method becoming more and more commonplace in relation to other types of voting or completion of official forms and having likely efficiency savings. There are remote voting procedures operating in at least one European country allowing the casting of a vote more than once by the same person, with only the final vote cast before close of poll counting. This is to provide for the possibility that an elector may be subject to intimidation when voting but would take a later opportunity to vote in private.
In the list of examples that might have been chosen, South Buckinghamshire in 2007 is a rather oddly specific choice. Plus which it’s very hard to locate those old voting trial documents online.
The usual assertion that online voting will be “popular”, without any context that online voting provably does not increase turnout.
I do like that there is at least some consideration given to security risks, but the idea that we should weigh “realistic concerns” about security against some vague notion of method popularity is odd. One should weigh the security risks of one type of voting against the security risks of another, and optimise for voting system integrity.
While being oddly specific about South Buckhamshire, the document is oddly vague about “at least one European country” – in fact there is only one country in the world that offers national Internet voting, Estonia, and it is only able to have multiple vote casting because it has a comprehensive nationwide system of digital ID, something which the Wales document doesn’t mention.
There is also no mention of the many countries that have had reports recommending against Internet voting (such as Canada) or countries that have withdrawn Internet voting due to security concerns (such as France).
(pages 19-20 of the main consultation document)
I don’t really have the energy to examine the electronic counting piece in detail. Basically what you need to know about electronic counting is that you MUST audit the counts because you cannot trust the counting machines (counting computers). Which, if you have a simple count anyway, means that you’ve generated more work and expense, not less. Electronic counting, with audits, only makes sense if you have a complicated count, and nevertheless distances the process of the election from direct public inspection and understanding.
As I have mentioned, a lot of the UK evidence from previous voting trials is now hard to locate online. But here are some nice clear statements from the UK Office of the Deputy Prime Minister (ODPM) in Implementation of Electronic Voting in the UK Technical Options Report circa 20031
A Comparison with Other Secure Transactions
It is useful to compare voting with other online transactions for which security is needed.
The most obvious comparison is with banking. Attacking an electronic voting system is unlikely to bring the immediate financial rewards that a successful attack on the banking system would, and thus some types of well-resourced attack are less likely. However, the likelihood of well-resourced attacks is still sufficiently high to be problematic.
The consequences of a successful attack are very different with electronic voting, than with banking, though. Banks can, and do, take a financial analysis of how much loss they can stand and insure against such losses. It may be that a political decision could be taken that the loss of a certain percentage of votes is acceptable, but in the absence of such a decision, security appropriate for banking cannot be considered sufficient for electronic voting. Banks have also maintained confidence in the face of repeated losses through computer crime by covering up the cause of those losses. It is inconceivable that, in the event of a successful attack on electronic voting, such a cover-up would be acceptable to the electorate if subsequently disclosed. In a similar vein, individuals can be, and are, compensated for financial losses due to disruption/failures/hacking of online banking. It is not easy to see how there could be equivalent compensation for disruption/failures/hacking of an individuals vote, even if somehow it was discovered which individuals were affected (which might not be possible with some sorts of disruption).
Another issue is anonymity: electronic voting differs from the aforementioned applications due to the fact that, in addition to the requirements for accuracy and privacy, there is the mandated necessity to provide … anonymity. In other words, banking applications can (in fact must) allow tracking back to the user of the system, but the [electronic voting system] must ensure that such tracking is impossible. (Mercuri, 2001, pp8-9).
Electronic voting also differs from financial transactions in that the risk that an election delayed by a few days will have a different result is unacceptably high. By contrast substantial financial transactions between two willing partners usually can be conducted a few days later if there are problems with ecommerce applications, since such transactions are rarely conducted on a whim.
The Mercuri citation above is to
Mercuri, Rebecca, 2001 Electronic Vote Tabulation: Checks and Balances PhD thesis, University of Pennsylvania.
1 From Paper Vote Canada blog post electronic voting in the UK – technical report, September 17, 2004. As the OPDM site is no longer available, a 31 July 2003 version from the Internet Archive is linked above.