Verifying Properties of Binarized Deep Neural Networks
This work addresses the challenge of ensuring reliability in deep learning for applications like image classification, though it is incremental as it builds on existing SAT solver techniques for a specific network type.
The paper tackles the problem of verifying properties of Binarized Neural Networks by proposing an exact Boolean encoding and using SAT solvers with a counterexample-guided search, demonstrating scalability to medium-size networks for robustness to adversarial perturbations in image classification tasks.
Understanding properties of deep neural networks is an important challenge in deep learning. In this paper, we take a step in this direction by proposing a rigorous way of verifying properties of a popular class of neural networks, Binarized Neural Networks, using the well-developed means of Boolean satisfiability. Our main contribution is a construction that creates a representation of a binarized neural network as a Boolean formula. Our encoding is the first exact Boolean representation of a deep neural network. Using this encoding, we leverage the power of modern SAT solvers along with a proposed counterexample-guided search procedure to verify various properties of these networks. A particular focus will be on the critical property of robustness to adversarial perturbations. For this property, our experimental results demonstrate that our approach scales to medium-size deep neural networks used in image classification tasks. To the best of our knowledge, this is the first work on verifying properties of deep neural networks using an exact Boolean encoding of the network.