Symbolic Execution for Deep Neural Networks

Abstract

Deep Neural Networks (DNN) are increasingly used in a variety of applications, many of them with substantial safety and security concerns. This paper introduces DeepCheck, a new approach for validating DNNs based on core ideas from program analysis, specifically from symbolic execution. The idea is to translate a DNN into an imperative program, thereby enabling program analysis to assist with DNN validation. A basic translation however creates programs that are very complex to analyze. DeepCheck introduces novel techniques for lightweight symbolic analysis of DNNs and applies them in the context of image classification to address two challenging problems in DNN analysis: 1) identification of important pixels (for attribution and adversarial generation); and 2) creation of 1-pixel and 2-pixel attacks. Experimental results using the MNIST data-set show that DeepCheck's lightweight symbolic analysis provides a valuable tool for DNN validation.

References in corpus (4)

• Axiomatic Attribution for Deep Networks
• DeepXplore: Automated Whitebox Testing of Deep Learning Systems
• Not Just a Black Box: Learning Important Features Through Propagating Activation Differences
• DeepSafe: A Data-driven Approach for Checking Adversarial Robustness in Neural Networks

Cited by in corpus (15)

• Testing Deep Neural Networks
• Machine Learning Testing: Survey, Landscapes and Horizons
• Do the Machine Learning Models on a Crowd Sourced Platform Exhibit Bias? An Empirical Study on Model Fairness
• A Survey of Safety and Trustworthiness of Deep Neural Networks: Verification, Testing, Adversarial Attack and Defence, and Interpretability
• Automated Test Generation to Detect Individual Discrimination in AI Models
• Metamorphic Testing for Object Detection Systems
• DeepSearch: A Simple and Effective Blackbox Attack for Deep Neural Networks
• DeepFault: Fault Localization for Deep Neural Networks
• Abstraction and Symbolic Execution of Deep Neural Networks with Bayesian Approximation of Hidden Features
• Explaining Image Classifiers using Statistical Fault Localization
• DeepSmartFuzzer: Reward Guided Test Generation For Deep Learning
• Reliability Validation of Learning Enabled Vehicle Tracking
• Efficient Adversarial Input Generation via Neural Net Patching
• Incremental Verification of Fixed-Point Implementations of Neural Networks
• Symbolic Techniques for Deep Learning: Challenges and Opportunities