There is a lot of great work getting done in the privacy space. The majority of this work focusses on the wallet / dapp, transaction layer, or chain layer. These are all very important from a data shielding standpoint, but it privacy breaks down at the RPC layer. Even if you are using Helios locally, it still must connect to RPC to validate state with eth_getProof for state calls. We end up leaving a data trail with a few, very trusted, centralized providers.

Providers like Alchemy, Quicknode, and Infura have long been criticized as centralized trusted RPC providers. Connecting to the RPC service leaks data in multiple ways, User accounts (emails, credit card payments, project IDs), traffic leakage (IP addresses from connecting from the wallet), payload leakage (method, parameters, and call data). While these providers don't sell / publish this data, we must trust they have diligence in protecting this data. Even if they don't sell it, this data can be mined and correlated to understand correlations between wallet, IP, and call patterns.

Projects like Kohaku are doing great work in a "fat" privacy wallet. It is important that a privacy first wallet come with a full privacy stack. Asking the user to install a tor client, have a local client running to connect to, or otherwise setup infrastructure for privacy focused transactions is too much to expected of a user. This project focuses primarily on the RPC (or "dial tone") part of the stack.

This project provides a way to get anonymous RPC. It can be thought of as dial tone from a payphone. No Credit Cards needed, pay with stables, the connection is anonymized via onion routing. Even if the gateway is connecting to a cloud provider, it doesn't leak location data.

1. The wasm based helios library is behind the repo head and did not support base-sepolia (or many OpStack testnets). I had to recompile the wasm library as a vendored library. This helps keep non-state calls local to the wallet.

2. The x402 facilitator gateway is a single point of failure and suffers from intermittent failures.

3. Getting a bundled tor client was a challenge. It was a solved via compiling Arti to multiple architectures and deploying as a managed service within an Electron app.