Towards Verifiable AI with Lightweight Cryptographic Proofs of Inference
Mar 19, 2026·
,,,,,,·
0 min read
Pranay Anchuri
Matteo Campanelli
Paul Cesaretti
Rosario Gennaro
Tushar M. Jois
Hasan S. Kayman
Tugce Ozdemir
Abstract
We address the challenge of verifying AI inference for cloud-deployed models. Rather than employing full cryptographic proofs — which impose prohibitive computational overhead — we present a verification framework and protocol that replaces full cryptographic proofs with a lightweight, sampling-based approach. The method uses Merkle-tree vector commitments to record execution traces, selectively opening entries along randomly sampled paths. Evaluated on ResNet-18 and Llama-2-7B, the approach reduces proving times from the order of minutes to the order of milliseconds.
Type
Publication
In IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) 2026
Authors
Senior Research Scientist
Pranay Anchuri is a Senior Research Scientist at Offchain Labs. His research spans
blockchain protocols, verifiable computation, and machine learning applied to
decentralized systems. He has published at top venues including KDD, JMLR, and ICDM,
and is an inventor on seven US patents. He holds a PhD in Computer
Science from Rensselaer Polytechnic Institute.