Ethereum co-founder Vitalik Buterin published the third installment of his cryptographic obfuscation series on Friday, examining a technique called "local mixing" that discards lattices and elliptic curves in favor of ideas drawn from hash function design.

The new post follows Part I, published on June 29, and the diamond iO writeup that followed on July 28, both of which centered on lattice-heavy constructions — the mathematical family that also underpins the post-quantum encryption standards NIST finalized in 2024. Local mixing, by contrast, is "a totally different way of doing cryptography," Buterin wrote in his latest post. The design contains no elliptic curves, no prime factorization, and no lattices anywhere, he noted. According to Buterin, the closest relative to local mixing is symmetric cryptography, the discipline behind everyday encryption and hashing. That is a battle-tested lineage: symmetric algorithms such as AES and SHA-2 remain in wide use and have withstood decades of open cryptanalysis, in contrast to constructions whose security rests on newer and less-tested assumptions.

Junk gates and a circuit run in reverse

The local mixing process begins with a circuit built from logic gates such as XOR, AND, and NOT. That circuit then passes through a pipeline that keeps the output identical while scrubbing away any trace of the internal logic. Buterin described the sequence: the original circuit is made reversible, then hardened, gadgetized, mixed, and finally obfuscated.

During the mixing stage, junk gates are scattered throughout the circuit, the arrangement is shuffled, and small blocks are swapped out for different gates that compute the same thing. Mixing alone is not enough, which is why the other steps matter, as they perform most of the work. Reversibility comes first because it makes the rest of the pipeline possible: Buterin explains that a reversible gate can be rewritten as an arbitrary pile of other reversible gates with matching behavior, something that is harder to achieve with an AND or an OR gate.

A bet built on past failures

Buterin nonetheless characterized local mixing as a "wild and risky bet," writing that it sits on "a graveyard of failed attempts at white-box cryptography." White-box cryptography asks for secrets to remain hidden inside software that an attacker can inspect and modify end to end — the threat model behind digital rights management and mobile payment code — and published schemes of that kind have repeatedly been broken, while ordinary symmetric cryptography sidesteps the problem by keeping algorithms public and secrets in keys held by users. Buterin pointed out, however, that the authors of local mixing say that more effort, along with a willingness to accept higher overhead, could make the idea hold.

One proposed shortcut is artificial intelligence, which could compress the three decades hash functions needed to mature into a span of a few years. That maturation happened in the open, through public contests such as NIST's SHA-3 competition, which ran from 2007 to 2012, and years of attempted breaks by the wider cryptanalysis community. The approach does move away from lattice-based routes, where the trade-off involved security assumptions.

Buterin called obfuscation the "final boss of cryptography" in his June publication, and "the final frontier of cryptography" in the latest installment. The most rigorous constructions, he said, carry "literally galactic" runtimes — longer than the lifetime of the universe.

Why obfuscation is worth the trouble

Obfuscation allows a program to be turned into an encrypted version that still runs on ordinary inputs, returning ordinary outputs while hiding the underlying code. In its strictest formal version, indistinguishability obfuscation, any two programs computing the same function should produce obfuscated forms that cannot be told apart. When paired with a blockchain, it gets close to a "trustless trusted third party," enabling applications such as private, collusion-resistant voting with no M-of-N committee to trust.