Inception

Hey hey, a couple of busy weeks behind us. Busy in a productive way, ofc. Let's dive into it!

Setup tweak

For the whole duration of the cohort, I was also working full-time and trying to carve out a couple of hours in my day for reading, analysis and prototyping. Over the past couple of weeks, I also needed to travel for work. To conclude and summarize the work that was left, I needed to come up with a more autonomous setup. I ended up dusting off my Raspberry Pi and setting up a Hermes agent with an Ollama subscription to be able to bypass the damn security guardrails.

Inception

The more autonomous setup ended up paying off pretty quickly. Across the last three weeks, the research turned into a lot more concrete implementation work: a full ml-dsa-44-commit/v0 account profile, from Python and TS implementations and deterministic vectors through an ERC-7913 signer, frame adapter, trusted-key registry, Solidity tests and e2e coverage. I also built StealthKeyExchange.sol with a Lean model, generated vectors, Solidity tests and TS client support, while reconciling the meta-address layout and version handling with the commitment scheme. On the spending side, I hardened verification around gas exhaustion, verifier errors, registry replacement and malformed ERC-7913 outputs. Along the way, a bunch of deterministic test infrastructure got fixed, and the docs/wiki were refreshed to better separate the key-exchange/ERC work from the still-speculative spending research.

Flexibility in mind

Looking at recently emerging tools for collaborative research, like sig.golf[1], it became clear to me that the proposed scheme MUST be flexible enough to keep the key exchange settled on ML-KEM while leaving the spending part to account abstraction.

Draft ethresear.ch post

With all things in mind, and with the pressing deadline to submit a presentation for a talk, I decided to start drafting a post with a summary of my research. I ended up with two schemes in total. The first one would use a secp256k1 spending path for immediate issue negotiation. The second one would remain unfinished so it can fully utilize all improvements coming in Hegota.

Positive MEV frontrun

It turned out that somebody was already working on adapting the stealth-address protocol for post-quantum times ahead of us. Not surprisingly, we quickly found consensus that, apart from the k1 spending path, there are too many moving parts to set things in stone. The thread with my comments is available here[2].

Reflection

Frankly, I'm extremely happy with how things played out. Given the amount of time I dedicated to research (not much), it's great to have more external validation for both the problem set and the solution. I'm planning to continue going through the post in detail and take a deeper look at the differences and common ground.

Remaining weeks

My next biggest priority is figuring out how to fit the existing research with my findings, and mapping which parts are ready to be wrapped up and proposed and what remains to be explored. The repo and documentation page for sure deserve some cleanup and clarifications.

Excited to see some of you in person!