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

Abstract

Despite the improved accuracy of deep neural networks, the discovery of adversarial examples has raised serious safety concerns. In this paper, we study two variants of pointwise robustness, the maximum safe radius problem, which for a given input sample computes the minimum distance to an adversarial example, and the feature robustness problem, which aims to quantify the robustness of individual features to adversarial perturbations. We demonstrate that, under the assumption of Lipschitz continuity, both problems can be approximated using finite optimisation by discretising the input space, and the approximation has provable guarantees, i.e., the error is bounded. We then show that the resulting optimisation problems can be reduced to the solution of two-player turn-based games, where the first player selects features and the second perturbs the image within the feature. While the second player aims to minimise the distance to an adversarial example, depending on the optimisation objective the first player can be cooperative or competitive. We employ an anytime approach to solve the games, in the sense of approximating the value of a game by monotonically improving its upper and lower bounds. The Monte Carlo tree search algorithm is applied to compute upper bounds for both games, and the Admissible A* and the Alpha-Beta Pruning algorithms are, respectively, used to compute lower bounds for the maximum safety radius and feature robustness games. When working on the upper bound of the maximum safe radius problem, our tool demonstrates competitive performance against existing adversarial example crafting algorithms. Furthermore, we show how our framework can be deployed to evaluate pointwise robustness of neural networks in safety-critical applications such as traffic sign recognition in self-driving cars.

References in corpus (7)

• End to End Learning for Self-Driving Cars
• Provable defenses against adversarial examples via the convex outer adversarial polytope
• Efficient Neural Network Robustness Certification with General Activation Functions
• Provably Minimally-Distorted Adversarial Examples
• Concolic Testing for Deep Neural Networks
• Global Robustness Evaluation of Deep Neural Networks with Provable Guarantees for the $L_0$ Norm
• Pixel-wise Segmentation of Street with Neural Networks

Cited by in corpus (30)

• Testing Deep Neural Networks
• Verification of Neural Network Behaviour: Formal Guarantees for Power System Applications
• How to Certify Machine Learning Based Safety-critical Systems? A Systematic Literature Review
• Analyzing Deep Neural Networks with Symbolic Propagation: Towards Higher Precision and Faster Verification
• Testing and verification of neural-network-based safety-critical control software: A systematic literature review
• A Survey of Safety and Trustworthiness of Deep Neural Networks: Verification, Testing, Adversarial Attack and Defence, and Interpretability
• A Safety Framework for Critical Systems Utilising Deep Neural Networks
• Chess AI: Competing Paradigms for Machine Intelligence
• The Vulnerability of Semantic Segmentation Networks to Adversarial Attacks in Autonomous Driving: Enhancing Extensive Environment Sensing
• A Survey on Verification and Validation, Testing and Evaluations of Neurosymbolic Artificial Intelligence
• The RFML Ecosystem: A Look at the Unique Challenges of Applying Deep Learning to Radio Frequency Applications
• Global Robustness Evaluation of Deep Neural Networks with Provable Guarantees for the $L_0$ Norm
• Evading classifiers in discrete domains with provable optimality guarantees
• The Feasibility and Inevitability of Stealth Attacks
• Adversarial Examples on Object Recognition: A Comprehensive Survey
• Sparse Adversarial Video Attacks with Spatial Transformations
• Fooling Object Detectors: Adversarial Attacks by Half-Neighbor Masks
• Towards the Quantification of Safety Risks in Deep Neural Networks
• Perturbation-based Self-supervised Attention for Attention Bias in Text Classification
• Formal Verification of Robustness and Resilience of Learning-Enabled State Estimation Systems
• DeepFault: Fault Localization for Deep Neural Networks
• Adversarial Robustness of Deep Learning: Theory, Algorithms, and Applications
• Abstraction and Symbolic Execution of Deep Neural Networks with Bayesian Approximation of Hidden Features
• Formal methods and software engineering for DL. Security, safety and productivity for DL systems development
• Reliability Validation of Learning Enabled Vehicle Tracking
• The Care Label Concept: A Certification Suite for Trustworthy and Resource-Aware Machine Learning
• BDD4BNN: A BDD-based Quantitative Analysis Framework for Binarized Neural Networks
• A Roadmap for Robust End-to-End Alignment
• NeVer 2.0: Learning, Verification and Repair of Deep Neural Networks
• Robustness Guarantees for Deep Neural Networks on Videos