Machina iO

Practical indistinguishability obfuscation. Roadmap through Q2 2027: blind PRF over key-homomorphic encodings, ≥64-bit obfuscation, SNARK verification, witness encryption.

Research

Roadmap Completion: 25%

Lattice

key-homomorphic encodings

FHE

GGH15

Now

FHE multiplication over key-homomorphic encodings

Implement FHE multiplication over key-homomorphic encodings. Foundation for blind PRF; unlocks predicate encryption / LFE.

Q1 2026 · The FHE multiplication circuit compatible with key-homomorphic encodings has been implemented. A further optimization and the evaluation of the circuit over the encodings is underway.

A paper on a new lookup table evaluation method over key-homomorphic encodings and low-depth modulo operation circuits

Publish a paper detailing a new method for evaluating lookup tables over key-homomorphic encodings, along with low-depth modulo operation circuits.

Q2 2026 · The method and the circuit have been implemented, and the paper is in the process of being written.

Noise refreshing with a dummy blind PRF

Implement noise refreshing of GGH15 encodings with replaceable dummy blind PRF. Confirm parameter growth is polylogarithmic.

Q2 2026 · The noise refreshing with a dummy blind PRF has not been implemented yet, but the high-level design has been finalized.

Blind PRF over key-homomorphic encodings

Implement a circuit compatible with key-homomorphic encodings that simulates a PRF without revealing key or output. The circuit will be used to replace the dummy blind PRF in the noise refreshing construction.

Q2–Q3 2026 · The circuit is expected to be implementable by combining the gadgets implemented in Q1 2026, although further efficiency improvements and careful parameter tuning are still needed.

Diamond iO 2

Implement the next version of the diamond iO, which will support a larger input size without compromising efficiency. A paper describing the construction and its security proof will be published as well.

Q3 2026

Devcon 2026: obfuscation for nontrivial input size

Demo and presentation: the first practical-performance iO for nontrivial input size (e.g., 64 input bits), or its concrete estimation of the required hardware costs and performance.

Q3 2026

Later

SNARK verification over key-homomorphic encodings

Implement a SNARK verification circuit over key-homomorphic encodings; PV vs DV scheme selection (or new design) plus an Argo-style garbled-circuit path.

Q4 2026 – Q1 2027

Witness encryption

Obfuscation that releases a hardcoded message iff a SNARK proof verifies, unlocking trustless bridges, encrypted mempools, one-time programs, and time-lock encryption.

Q1 – Q2 2027

Description

✓Improve efficiency of key-homomorphic encodings to enable blind PRF and noise refreshing
✓Diamond iO 2, supporting larger input sizes without compromising efficiency
✓SNARK verification over key-homomorphic encodings (Q4 2026 – Q1 2027)
✓Witness encryption as the first end-user-relevant deliverable (Q1 – Q2 2027)

Deliverables

✓Implementation of FHE multiplication + blind FHE over key-homomorphic encodings, along with noise refreshing
✓Implementation of obfuscation with a 64-bit input size or its concrete estimation of the required hardware costs and performance
✓Papers on the new lookup table evaluation method and low-depth circuit in Q2 and the next diamond iO construction in Q3
✓SNARK verification circuit and prover (Q4 2026 – Q1 2027), then witness encryption with a demo application (Q1 – Q2 2027)

Impact

✓First practical-performance iO for nontrivial input size
✓First witness encryption for general NP, enabling trustless bridges, encrypted mempools, and time-lock encryption
✓Foundation for verifiable FHE and obfuscation for conditional FHE decryption (out of scope this cycle)