Signal New taxonomy documents non-adversarial failure modes in optical-scan voting systems

Summary

Researchers presented a taxonomy of non-adversarial failure modes found in optical-scan voting systems and the surrounding ecosystem of people, processes and technology, illustrated with documented real-world cases. Unlike much prior work, which has focused on adversarial threats to voting systems, this paper notes that many jurisdictions have not fully prepared for the possibility of ordinary tabulator error. The researchers also observe that stakeholders tend to find real-world accounts of equipment and procedural failures more convincing than hypothetical attack scenarios. The taxonomy sorts failures into four categories, recording votes on paper, reading votes from paper, combining readings into a reported outcome, and testing and verification. The researchers mapped existing verification mechanisms onto this taxonomy and identified gaps that no paper-based audit can detect or correct. Failures that undermine the trustworthiness of the paper trail itself, such as giving voters the wrong ballot style or misusing ballot-marking devices, stood out as particularly significant.