arXiv (Cornell University) · 2018 · 188 citations · 0 references
ComputingComputer ScienceSystems EngineeringFormal MethodsMachine LearningSafety EngineeringDeductive VerificationNetwork RobustnessNeural NetworksSafety-critical SystemsAdversarial RobustnessFormal Methods For SafetyFormal Safety VerificationMalware DetectionSafety Critical System AnalysisSafety PropertiesSafety Property VerificationSafety Verification
Neural networks are increasingly used in safety‑critical domains such as autonomous driving and aircraft collision avoidance, yet they can mispredict on minor perturbations, leading to potentially disastrous outcomes, and current analysis methods struggle to scale or reliably produce counterexamples. The paper aims to provide a formal analysis system that rigorously checks neural networks for safety property violations, such as robustness against bounded adversarial perturbations. We propose an efficient method that rigorously checks safety properties and produces concrete counterexamples for networks up to ten times larger than those supported by existing techniques, outperforming prior approaches by orders of magnitude. By tightly bounding network outputs over input ranges, our method enhances explainability and can inform training of more robust neural networks.
Neural networks are increasingly deployed in real-world safety-critical domains such as autonomous driving, aircraft collision avoidance, and malware detection. However, these networks have been shown to often mispredict on inputs with minor adversarial or even accidental perturbations. Consequences of such errors can be disastrous and even potentially fatal as shown by the recent Tesla autopilot crash. Thus, there is an urgent need for formal analysis systems that can rigorously check neural networks for violations of different safety properties such as robustness against adversarial perturbations within a certain L-norm of a given image. An effective safety analysis system for a neural network must be able to either ensure that a safety property is satisfied by the network or find a counterexample, i.e., an input for which the network will violate the property. Unfortunately, most existing techniques for performing such analysis struggle to scale beyond very small networks and the ones that can scale to larger networks suffer from high false positives and cannot produce concrete counterexamples in case of a property violation. In this paper, we present a new efficient approach for rigorously checking different safety properties of neural networks that significantly outperforms existing approaches by multiple orders of magnitude. Our approach can check different safety properties and find concrete counterexamples for networks that are 10x larger than the ones supported by existing analysis techniques. We believe that our approach to estimating tight output bounds of a network for a given input range can also help improve the explainability of neural networks and guide the training process of more robust neural networks.