Safety Verification of Deep Neural Networks. (arXiv:1610.06940v1 [cs.AI])

Deep neural networks have achieved impressive experimental results in image classification, but can surprisingly be unstable with respect to adversarial perturbations, that is, minimal changes to the input image that cause the network to misclassify it. With potential applications including perception modules and end-to-end controllers for self-driving cars, this raises concerns about their safety. We develop the first SMT-based automated verification framework for feed-forward multi-layer neural networks that works directly with the code of the network, exploring it layer by layer. We define safety for a region around a data point in a given layer by requiring that all points in the region are assigned the same class label. Working with a notion of a manipulation, a mapping between points that intuitively corresponds to a modification of an image, we employ discretisation to enable exhaustive search of the region. Our method can guarantee that adversarial examples are found for the given region and set of manipulations. If found, adversarial examples can be shown to human testers and/or used to fine-tune the network, and otherwise the network is declared safe for the given parameters. We implement the techniques using Z3 and evaluate them on state-of-the-art networks, including regularised and deep learning networks.

from cs.AI updates on arXiv.org http://ift.tt/2eLPpVi
via IFTTT