Internet voting filter bubbles

From a Canadian perspective, there are basically three groups that examine Internet voting:

  • social scientists that examine people’s attitudes, feelings and behaviours associated with Internet voting
  • staff at municipalities that have chosen Internet voting and see it as just another digital service to offer, and the vendors they procure Internet voting from
  • computer scientists that examine Internet voting from the perspective of requirements and threat risk assessment

These three communities basically don’t interact.  The social scientists cite one another.  The municipal staff and vendors reference other municipalities and vendor analysis.  The computer scientists cite one another.  This gives three basically different filtered world views.

  • The social science perspective indicates some level of popularity of Internet voting either conceptually or in practice, and associated levels of satisfaction.  It also documents the expectations of turnout (high) and the reality of turnout (no change).  Additionally and unfortunately it sometimes reports on perceptions of security, which are meaningless.  It doesn’t matter how safe you feel jumping off a cliff, the same thing will still happen at the bottom when you encounter reality.
  • At best, municipalities approach Internet voting from a digital services perspective, and do the standard things one does for a transactional service, including security buzzwords like firewalls and encryption, obtaining vendor assurances, and contracting confidential security assessments.  One of their primary sources of technical information is the vendors themselves.  Two issues are that Internet voting is not a standard transactional service, and that vendors have literally millions of dollars in sales at stake.
  • Computer scientists look at the requirements for voting systems, e.g. the Computer Technologists’ Statement on Internet Voting.  When they evaluate real Internet voting systems against those requirements, they always find that current systems cannot meet the requirements.  In order to provide the best security assessment of the real systems, they seek the ability to conduct truly independent and public security assessments of the technology being used (this is almost always denied).  They also assess the full spectrum of potential risks against a system.  That includes technical risks and non-technical risks.  An often overlooked risk is the risk of coercion when voting no longer takes place in private in a supervised location (the polling place).  They also examine techniques used by very sophisticated attackers, as well as very basic but successful techniques (e.g. phishing) and the risk of insider attacks.  For a service where there is no way for the end user to verify their intended result (due to the combination of secret ballot and coercion avoidance), the inevitable conclusion is that there are no adequate risk mitigation measures.

So the answer you get about Internet voting depends on which community you ask.  If you ask social scientists, it’s popular.  If you ask municipalities that have implemented it, they assure you that everything is going fine.  If you ask computer scientists, they will tell you that it is not a regular transactional digital service, and that using Internet voting introduces catastrophic risk.

You can get a pretty easy indication of which community is talking by looking for language clues.  If the discussion is around popularity, it’s probably a social science analysis.  If the discussion is around firewalls and encryption and security assurances, it’s probably municipalities.  If the discussion is around risks, it’s probably computer scientists.

It may seem odd that computer scientists would speak in less technical language, but that’s because specific technical measures are much less important than a system-wide requirements and threat analysis, particularly in an environment including home computing devices and non-technical online service users.

The result of having these different communities means that basically only consultations that include the computer science community recommend against voting using computers, which may be an unexpected outcome.  But it is the outcome of any serious consultation, including e.g. New Brunswick, Nova Scotia, Quebec, Ontario, British Columbia, the Government of Canada, and the Government of Australia.

The Ontario municipal association AMCTO is holding a 2017 event for municipal clerks, featuring a session about the security of Internet voting.  The presenters will be

  • a clerk from a municipality that has approved Internet voting
  • an Internet voting vendor representative
  • a second Internet voting vendor representative

I leave it to you to conclude which filter bubble will be in operation.