exttt{Range-Arithmetic}: Verifiable Deep Learning Inference on an Untrusted Party

2025-05-23

AI Summary

Efficient verifiability of outsourced DNN inference in decentralized machine learning remains challenging, particularly due to non-arithmetic operations (e.g., ReLU, fixed-point rounding) incompatible with arithmetic zero-knowledge proof systems over finite fields. Method: We propose the first purely arithmetic compilation scheme supporting ReLU and fixed-point rounding—avoiding Boolean encoding and high-degree polynomials—by integrating range proofs with the sum-check protocol. We design arithmetic circuits for fixed-point matrix multiplication and arithmeticized activation functions, enabling end-to-end verifiable inference. Contribution/Results: Our approach reduces verification cost, prover computation overhead, and communication volume significantly compared to state-of-the-art schemes, while preserving model accuracy. Crucially, it losslessly compiles nonlinear, non-arithmetic operations into low-degree arithmetic circuits, enabling practical, lightweight blockchain clients to securely invoke DNN services.

Technology Category

• Machine Learning: Hardware-aware ML
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Verifiable computing (VC) has gained prominence in decentralized machine learning systems, where resource-intensive tasks like deep neural network (DNN) inference are offloaded to external participants due to blockchain limitations. This creates a need to verify the correctness of outsourced computations without re-execution. We propose exttt{Range-Arithmetic}, a novel framework for efficient and verifiable DNN inference that transforms non-arithmetic operations, such as rounding after fixed-point matrix multiplication and ReLU, into arithmetic steps verifiable using sum-check protocols and concatenated range proofs. Our approach avoids the complexity of Boolean encoding, high-degree polynomials, and large lookup tables while remaining compatible with finite-field-based proof systems. Experimental results show that our method not only matches the performance of existing approaches, but also reduces the computational cost of verifying the results, the computational effort required from the untrusted party performing the DNN inference, and the communication overhead between the two sides.

Problem

• Verifying correctness of outsourced DNN inference without re-execution
• Transforming non-arithmetic DNN operations into verifiable arithmetic steps
• Reducing computational cost and communication overhead in verifiable computing

Innovation

• Transforms non-arithmetic operations into verifiable arithmetic steps
• Uses sum-check protocols and concatenated range proofs
• Reduces computational cost and communication overhead