AI verification

• AI verification
  • Certification for Neural network
    • Overview
    • AI Explainability 360
    • Adversarial Robustness 360 Toolbox (ART)
    • PROVEN: Verifying Robustness of Neural Networks with a Probabilistic Approach
    • CROWN: A Neural Network Verification Framework
    • Efficient Neural Network Robustness Certification with General Activation Functions
    • Certified Adversarial Robustness via Randomized Smoothing
    • CNN-Cert: An Efficient Framework for Certifying Robustness of Convolutional Neural Networks
    • THE ROBUSTNESS OF NEURAL NETWORKS: AN EXTREME VALUE THEORY APPROACH
    • Robustness Verification of Tree-based Models
    • Efficient Formal Safety Analysis of Neural Networks
    • A Game-Based Approximate Verification of Deep Neural Networks with Provable Guarantees
    • DeepGO: Reachability Analysis of Deep Neural Networks with Provable Guarantees
  • Attack and Robustness
    • Robust-and-Explainable-machine-learning
    • Is Robustness the Cost of Accuracy? A Comprehensive Study on the Robustness of 18 Deep Image Classification Models
    • AutoZOOM: Autoencoder-based Zeroth Order Optimization Method for Attacking Black-box Neural Networks
  • Related Topics

中文版

AI verification is a special point to view the booming research and applications of AI, especially deep learning.

Considering the wide and rapidly increasing use of AI among all kinds of industry, a professional inspection and certification from third-party organizations becomes important for both customers and the whole society.

Data plays an unprecedented role in deep learning. It also unavoidably involves public issues, such as privacy, ethics, risk and safety.

Verification for Neural network

Overview

IBM has done a great work on related topics. MIT and Oxford also have some related research.

AI Explainability 360

https://github.com/IBM/AIX360

Adversarial Robustness 360 Toolbox (ART)

https://github.com/IBM/adversarial-robustness-toolbox

中文版

PROVEN: Verifying Robustness of Neural Networks with a Probabilistic Approach

http://proceedings.mlr.press/v97/weng19a/weng19a.pdf

CROWN: A Neural Network Verification Framework

https://github.com/IBM/CROWN-Robustness-Certification

Efficient Neural Network Robustness Certification with General Activation Functions

https://arxiv.org/pdf/1811.00866.pdf

Certified Adversarial Robustness via Randomized Smoothing

https://github.com/locuslab/smoothing

https://arxiv.org/abs/1902.02918

medium blog and video

CNN-Cert: An Efficient Framework for Certifying Robustness of Convolutional Neural Networks

https://github.com/IBM/CNN-Cert

https://www.aaai.org/ojs/index.php/AAAI/article/view/4193

THE ROBUSTNESS OF NEURAL NETWORKS: AN EXTREME VALUE THEORY APPROACH

https://arxiv.org/pdf/1801.10578.pdf

Robustness Verification of Tree-based Models

https://arxiv.org/pdf/1906.03849.pdf

Efficient Formal Safety Analysis of Neural Networks

http://papers.nips.cc/paper/7873-efficient-formal-safety-analysis-of-neural-networks.pdf

A Game-Based Approximate Verification of Deep Neural Networks with Provable Guarantees

https://arxiv.org/pdf/1807.03571.pdf

https://github.com/TrustAI/DeepGame

DeepGO: Reachability Analysis of Deep Neural Networks with Provable Guarantees

https://arxiv.org/abs/1805.02242

https://github.com/TrustAI/DeepGO

Attack and Robustness

Robust-and-Explainable-machine-learning

https://github.com/dongyp13/Robust-and-Explainable-Machine-Learnings

Is Robustness the Cost of Accuracy? A Comprehensive Study on the Robustness of 18 Deep Image Classification Models

https://arxiv.org/abs/1808.01688

https://github.com/huanzhang12/Adversarial_Survey

AutoZOOM: Autoencoder-based Zeroth Order Optimization Method for Attacking Black-box Neural Networks

https://arxiv.org/abs/1805.11770

https://github.com/IBM/Autozoom-Attack

Related Topics

• AI safety
• Explainable/Interpretable AI (XAI)
• AI robustness
• GAN defense and attack
• Certification framework