Formal model-based validation for tally systems

• Authors: Dermot Cochran;Joseph R. Kiniry
• Venue: Vote-ID'13 Proceedings of the 4th international conference on E-Voting and Identity
• Year: 2013

Existing commercial and open source e-voting systems have horrifically poor testing frameworks. Most tally systems, for example, are tested by re-running all past elections and seeing if the new system gives the same answer as an older, perhaps erroneous, system did. This amounts to a few dozen system tests and, typically, few-to-no unit tests. These systems are used today in a dozen countries to determine the outcome of national elections. This state-of-affairs cannot continue because it calls into question the legitimacy of elections in major European and North American democracies. In this work, the ballot counting process for one of the most complex electoral schemes used in the world, Proportional Representation by Single Transferable Vote (PR-STV), is mechanically formally modeled. The purpose of such a formalization is to generate, using an algorithm of our design, a complete set of non-isomorphic test cases per electoral scheme, once and for all. Using such a system test suite, any digital election technology (proprietary or open source) can be rigorously evaluated for correctness. Doing so will vastly improve the confidence experts have--and can only improve the level of trust citizens have--in these digital elections systems.