The First Steps Toward Practical iO?
How Machina iO is trying to turn one of cryptography’s hardest problems into working code
In this week’s episode, Tarun and I spoke with Sora Suegami and Enrico Giudice from Machina iO, a team within the Ethereum Foundation’s PSE group, about their efforts to build usable indistinguishability obfuscation (iO) tools.
We’d covered iO once before in our episode with Rachel Lin, where we focused on the theoretical foundations and the potential of iO. This time, we explored the very early implementation work happening at Machina iO. I knew from the interview with Rachel and speaking with Tarun afterwards, that truly practical iO is still a long ways off. But it was cool to hear that there was a team within the EF at least attempting to take the first steps in this very uncharted territory.
In the interview, we were exploring a project still in the phase where multiple implementation paths are attempted, these paths are benchmarked and analysed, tradeoffs are decided on and new assumptions are created in order to push forward.
Key takeaways from Episode 384
Machina iO is a research team under the Ethereum Foundation’s PSE program exploring early implementations of indistinguishability obfuscation (iO). Their longer-term goal: make iO a viable replacement for trust assumptions in MPC and TEEs — turning it into a building block for next-generation privacy architectures.
Current iO constructions remain highly inefficient — obfuscating a 4-bit circuit can take over 6 hours and produce a 50 GB binary.
Sora and Enrico’s introduced us to new variants of assumptions like their all-product LWE and evasive LWE.
Their experiments borrow techniques like lookup table evaluation from modern ZK and FHE optimizations.
They see future iO systems overlapping with FHE VMs and conditional decryption use cases.
Despite the the slow generation time, Machina.io’s benchmarks provide rare empirical data for a field that’s been mostly theoretical until now.
In a way, this episode was a throwback to early zk-focused episodes circa 2018, where the theoretical research was just starting to be transformed into tangible codebases. Back then, research was moving fast, but the tooling still felt like it was years away from being viable (or testable). Nowadays in ZK, we have real benchmarks and strong fundamentals to work with. iO on the other hand is still somewhat in its primordial phase. In technical development cycles, I feel like this moment is one of the most thrilling, but also frustrating. Things are on the cusp of working, but there is still a long road ahead!
Be sure to check out Episode 384: the full episode page is here, or follow this thread for commentary.
Talk soon,
Anna
ECOSYSTEM ANNOUNCEMENT
We want to hear from you!
We’re trying to learn more about what topics you’d like us to cover, where our blind spots might be, and a bit about who our listeners are. We’ve put together a short form you can fill out anonymously. We’d love to hear from you and how you think we can make the Zero Knowledge Podcast better!