Shared Certificates for Neural Network Verification

TL;DR

This paper introduces shared certificates, a novel method that reuses verification proofs across multiple inputs to significantly reduce the computational cost of neural network verification without sacrificing accuracy.

Contribution

The paper presents the concept of shared certificates, enabling proof reuse in neural network verification, which is a novel approach to improve efficiency over traditional methods.

Findings

Shared certificates reduce verification time significantly.

Effective across various datasets and perturbation types.

Maintains verification precision while improving efficiency.

Abstract

Existing neural network verifiers compute a proof that each input is handled correctly under a given perturbation by propagating a symbolic abstraction of reachable values at each layer. This process is repeated from scratch independently for each input (e.g., image) and perturbation (e.g., rotation), leading to an expensive overall proof effort when handling an entire dataset. In this work, we introduce a new method for reducing this verification cost without losing precision based on a key insight that abstractions obtained at intermediate layers for different inputs and perturbations can overlap or contain each other. Leveraging our insight, we introduce the general concept of shared certificates, enabling proof effort reuse across multiple inputs to reduce overall verification costs. We perform an extensive experimental evaluation to demonstrate the effectiveness of shared certificates in reducing the verification cost on a range of datasets and attack specifications on image classifiers including the popular patch and geometric perturbations. We release our implementation at https://github.com/eth-sri/proof-sharing.