Using Zero-Knowledge Proofs to Verify Machine Learning Algorithms On-Chain

Blockchains have incentivized zero-knowledge proof research, creating modern proving systems w/ smaller memory footprints and faster proving & verification times — making it possible to verify certain small machine learning algorithms on-chain. https://t.co/NzvnXEX5BE

— Elena (@VirtualElena) April 5, 2023

ZKPs are useful for more than validity rollups and on-chain privacy. Recently, there have been notable efforts in putting larger and more complex models on-chain, leveraging the ability of ZKPs to succinctly represent and verify programs.

— Elena (@VirtualElena) April 5, 2023

There’s still work to be done, notably in representing larger models: accuracy may suffer in the process of converting a model’s architecture (layers, nodes, and activation functions), parameters, constraints, and matrix multiplication operations into a custom circuit.

— Elena (@VirtualElena) April 5, 2023

Still the space is robust and maturing quickly, with frameworks for verifying model authenticity, integrity, ingesting inputs from attested hardware, and doing decentralized inference (and training) on-chain.

— Elena (@VirtualElena) April 5, 2023

Thanks to @SuccinctJT, @danboneh, @guywuolletjr, @samrags_, @alive_eth, @cdixon, @eddylazzarin, @Tim_Roughgarden, @rhhackett, @Tim_Org for edits and/or inspiration and @jasonmorton, @LiaoPeiyuan, @tarunchitra, @BrianRetford, @daniel_d_kang, @theyisun, @AnnaRRose, @ModulusLabs,…

— Elena (@VirtualElena) April 5, 2023