Scaling Polyhedral Neural Network Verification on GPUs
This addresses the need for reliable neural network verification in safety-critical systems like autonomous driving and medical diagnosis, representing a strong specific gain.
The paper tackles the problem of scaling neural network verification for robustness against adversarial attacks, introducing GPUPoly which can verify a 1M neuron, 34-layer deep residual network in approximately 34.5 ms, significantly larger than previously possible.
Certifying the robustness of neural networks against adversarial attacks is essential to their reliable adoption in safety-critical systems such as autonomous driving and medical diagnosis. Unfortunately, state-of-the-art verifiers either do not scale to bigger networks or are too imprecise to prove robustness, limiting their practical adoption. In this work, we introduce GPUPoly, a scalable verifier that can prove the robustness of significantly larger deep neural networks than previously possible. The key technical insight behind GPUPoly is the design of custom, sound polyhedra algorithms for neural network verification on a GPU. Our algorithms leverage the available GPU parallelism and inherent sparsity of the underlying verification task. GPUPoly scales to large networks: for example, it can prove the robustness of a 1M neuron, 34-layer deep residual network in approximately 34.5 ms. We believe GPUPoly is a promising step towards practical verification of real-world neural networks.