Deep Learning Toolbox Interface for alpha-beta-CROWN Verifier

Verify robustness properties of PyTorch and ONNX Deep Neural Networks using the α,β-CROWN (alpha-beta-CROWN) Verifier

The Deep Learning Toolbox™ Interface for alpha-beta-CROWN Verifier enables verification of neural networks in ONNX and PyTorch formats, including computation of network bounds, robustness of classification networks against input perturbation, and generation of adversarial examples, i.e., imperceptible perturbations applied to inputs to the network that cause misclassification.

This interface provides access to the state-of-the-art formal verification algorithm, α,β-CROWN, winner of the annual neural network verification competition, VNN-COMP, each year from 2021 to 2025.

References:

[1] Verified Intelligence. Verified-Intelligence/Alpha-Beta-CROWN. 29 Jun. 2021, Python. GitHub, https://github.com/Verified-Intelligence/alpha-beta-CROWN.