Janis Erdmanis
Jul 5, 2025 | 1290 Words

Failure that Stings

I have always kept multiple options available and diversified my activities so that failure in one would not bring me down if it were to occur. I've had a few successes with this approach to life by maintaining my fitness, engaging in woodworking, which includes breaking logs into planks with a sawmill, installing solar panels, and setting up an NVR for my home. This time, however, my recent failure to get my conference paper into the main e-voting conference stings badly as the future looks grim.

After completing my PhD in condensed matter physics, I have been consistently working on the e-voting problem, exploring unconventional approaches to previously unaddressed issues. The problem I am interested in is reducing deployment complexity while maintaining voters' privacy and vote integrity. For a while, I was working on bringing a history tree-enabled bulletin board to the picture. This prevents votes from being removed from the server, where voters' clients serve as monitors to maintain consistency in a replication-free manner. In combination with previous work on DSA public key anonymisation, published by other prominent authors a decade ago, it enables fully centralised deployment.

I wrote a paper on the system and submitted it to the conference discussing various benefits of the approach. My mistake was to overreach by proposing a receipt-freeness mechanism that relied on authority withholding verifiability. This was based on the assumption that voters would not fear the consequences of following coercers' orders if those were far in the future, much like people often take debt. This, along with some implementation and protocol design mix-ups, was a strong basis for the rejection that I received from the reviewers.

It made me weep. One aspect that helped to deal with the grief was that I only invested 15 days in writing from start to finish, as I had already anticipated potential failure. I was more interested in presenting a prototype at JuliaCon 2024, which I did along with the AppBundler project.

The AppBundler project was created as part of the evoting project to resolve the challenges of deploying Julia GUI applications across all platforms. It was an outstanding success, and the talk now has over 1,000 views on YouTube! It also enabled me to get a six-month funding from NLNET to work exclusively on the project. Hence, it marks the first success of the activities I have undertaken as part of the PeaceFounder project.

The PeaceFounder project is significantly more complex, as it features full cryptographic implementations, an admin panel and a QML user client. I presented it at JuliaCon 2024 in the context of full-stack application development, as it encompasses all the necessary elements and could help expand on what is possible to achieve within the Julia ecosystem. The talk, however, had little to no impact on the community, and I got no interest or no collaborators.

This is when I realised the fundamental truth that ordinary people, even those with engineering backgrounds, are not typically interested in understanding voting security, leaving that to the experts. The experts that people trust are hopefully coming from academia and are involved in the e-voting problem space. Unfortunately, I am not one of those people whom people in that set would trust for their judgment.

I had an option of continuing to polish the PeaceFounder project, bringing it from alpha to beta stage, but it would be hard to gain an audience for it. Hence, I completely changed my focus to working on an academically relevant problem of combining individual verifiability with receipt-freeness, which was a strong basis for rejection of the paper from the conference and, to my understanding, is a central puzzle in e-voting research.

I employed an experimental approach by exploring what could be done with the exponential mix, my favourite primitive. What I love about it is how it hides the complexity of zero-knowledge proofs behind very understandable trust assumptions. The inputs and outputs can be interpreted as a physical braiding process, with secret keys representing individual threads in the braid. I also knew I needed to aim for everlasting privacy, as the quantum-hype machine had made stakeholders fear the potential of quantum computers.

The research was highly chaotic and involved convincing myself of false statements, then working from there as if they were true. This approach often opens unconventional reasoning paths and stimulates creativity, a benefit of working alone without needing to explain every step. The drawback is that I couldn't explain how I arrived at ideas without walking through all the incremental system changes that led to them. This chaotic process led me to explore tracker constructions, which initially failed. However, I incidentally came up with a deniable vote updating mechanism. That led me to focus on separating the voting device from the voting calculator until I had a breakthrough, where I became aware that physical information control can be used as a security mechanism for individual verifiability.

I distilled the tracker construction from the system and spent months painstakingly writing the paper. I kept it within the 20-page limit for work-in-progress submissions that the conference accepts. I am quite satisfied with the flow of the resulting paper https://eprint.iacr.org/2025/1186 and the arguments I had given for the need of a trapdoorless tracker construction. I also created neat figures illustrating the information flow and voters' interaction during the voting and optional verification.

I submitted the paper on May 15th to the major voting conference that takes place annually in Europe in October. On June 24th, I received the disappointing news that the paper was not among the selected ones. Two of the three reviewers recognised the value in the paper and graded it with weak acceptance, mainly pointing out only the absence of formal security definitions and game-theoretic proofs, but expressing confidence that such definitions do exist. The third reviewer made a weak attempt at understanding the paper, judging from their comments, and graded it with a strong rejection. That dashed my hope of securing a postdoc where I could formalise the system, explore extensions, and find potential collaborators with whom to bring it to life in a prototype.

I have been diligently pursuing the most impactful references for the system by sending emails to the authors for guidance and collaboration on formalising the security definitions and proofs, while also expressing my intention to pursue a postdoctoral position where I could further develop this work. So far, I have received replies from only two of the six individualised emails I sent two weeks ago, but no leads for active correspondence on the matter for now. There may still be hope, but it seems like I am in misery now.

It would have felt better if the system were not sound, and reviews would have opened my eyes about it. Instead, I'm left holding what I believe is a real contribution that the field won't engage with. Two out of three reviewers saw value, but that wasn't enough to make the work visible at the conference. The work is sound and novel, but that apparently isn't sufficient to secure some income to work on the problem or to find collaborators who could help me with formal definitions and proofs. Being right while professionally invisible is its own special kind of hell.

CC BY-SA 4.0 Janis Erdmanis. Last modified: July 05, 2025. Website built with Franklin.jl and the Julia programming language.