Formal Methods for Trustworthy Voting Systems : From Trusted Components to Reliable Software

Voting is prominently an important part of democratic societies, and its outcome may have a dramatic and broad impact on societal progress. Therefore, it is paramount that such a society has extensive trust in the electoral process, such that the system’s functioning is reliable and stable with respect to the expectations within society. Yet, with or without the use of modern technology, voting is full of algorithmic and security challenges, and the failure to address these challenges in a controlled manner may produce fundamental flaws in the voting system and potentially undermine critical societal aspects.
In this thesis, we argue for a development process of voting systems that is rooted in and assisted by formal methods that produce transparently checkable evidence for the guarantees that the final system should provide so that it can be deemed trustworthy. The goal of this thesis is to advance the state of the art in formal methods that allow to systematically develop trustworthy voting systems that can be provenly verified. In the literature, voting systems are modeled in the following four comparatively separable and distinguishable layers: (1) the physical layer, (2) the computational layer, (3) the election layer, and (4) the human layer. ... mehrCurrent research usually either mostly stays within one of those layers or lacks machine-checkable evidence, and consequently, trusted and understandable criteria often lack formally proven and checkable guarantees on software-level and vice versa.
The contributions in this work are formal methods that fill in the trust gap between the principal election layer and the computational layer by a reliable translation of trusted and understandable criteria into trustworthy software. Thereby, we enable that executable procedures can be formally traced back and understood by election experts without the need for inspection on code level, and trust can be preserved to the trustworthy system.

The works in this thesis all contribute to this end and consist in five distinct contributions, which are the following:

(I) a method for the generation of secure card-based communication schemes,
(II) a method for the synthesis of reliable tallying procedures,
(III) a method for the efficient verification of reliable tallying procedures,
(IV) a method for the computation of dependable election margins for reliable audits,
(V) a case study about the security verification of the GI voter-anonymization software.

These contributions span formal methods on illustrative examples for each of the three principal components, (1) voter-ballot box communication, (2) election method, and (3) election management, between the election layer and the computational layer.
Within the first component, the voter-ballot box communication channel, we build a bridge from the communication channel to the cryptography scheme by automatically generating secure card-based schemes from a small formal model with a parameterization of the desired security requirements. For the second component, the election method, we build a bridge from the election method to the tallying procedure by (1) automatically synthesizing a runnable tallying procedure from the desired requirements given as properties that capture the desired intuitions or regulations of fairness considerations, (2) automatically generating either comprehensible arguments or bounded proofs to compare tallying procedures based on user-definable fairness properties, and (3) automatically computing concrete election margins for a given tallying procedure, the collected ballots, and the computed election result, that enable efficient election audits. Finally, for the third and final component, the election management system, we perform a case study and apply state-of-the-art verification technology to a real-world e-voting system that has been used for the annual elections of the German Informatics Society (GI – “Gesellschaft für Informatik”) in 2019. The case study consists in the formal implementation-level security verification that the voter identities are securely anonymized and the voters’ passwords cannot be leaked.
The presented methods assist the systematic development and verification of provenly trustworthy voting systems across traditional layers, i.e., from the election layer to the computational layer. They all pursue the goal of making voting systems trustworthy by reliable and explainable formal requirements. We evaluate the devised methods on minimal card-based protocols that compute a secure AND function for two different decks of cards, a classical knock-out tournament and several Condorcet rules, various plurality, scoring, and Condorcet rules from the literature, the Danish national parliamentary elections in 2015, and a state-of-the-art electronic voting system that is used for the German Informatics Society’s annual elections in 2019 and following.