Zero-Knowledge Proof Based Verifiable Inference of Models

TL;DR

This paper presents a zero-knowledge proof framework that enables verification of AI model inference without revealing sensitive internal parameters, ensuring privacy and integrity in AI applications.

Contribution

It introduces a novel zero-knowledge proof system supporting complex neural network layers, with no trusted setup, and demonstrates its practical application on real AI models.

Findings

Supports both linear and nonlinear neural network layers.

Produces succinct, constant-size proofs using zkSNARKs.

Demonstrates efficiency and flexibility on real-world AI models.

Abstract

Recent advances in artificial intelligence (AI), particularly deep learning, have led to widespread adoption across various applications. Yet, a fundamental challenge persists: how can we verify the correctness of AI model inference when model owners cannot (or will not) reveal their parameters? These parameters represent enormous training costs and valuable intellectual property, making transparent verification difficult. In this paper, we introduce a zero-knowledge framework capable of verifying deep learning inference without exposing model internal parameters. Built on recursively composed zero-knowledge proofs and requiring no trusted setup, our framework supports both linear and nonlinear neural network layers, including matrix multiplication, normalization, softmax, and SiLU. Leveraging the Fiat-Shamir heuristic, we obtain a succinct non-interactive argument of knowledge (zkSNARK) with constant-size proofs. To demonstrate the practicality of our approach, we translate the DeepSeek model into a fully SNARK-verifiable version named ZK-DeepSeek and show experimentally that our framework delivers both efficiency and flexibility in real-world AI verification workloads.