Incorrect by Construction: Fine Tuning Neural Networks for Guaranteed Performance on Finite Sets of Examples

08/03/2020

∙

by   Ivan Papusha, et al.

∙

0

∙

share

There is great interest in using formal methods to guarantee the reliability of deep neural networks. However, these techniques may also be used to implant carefully selected input-output pairs. We present initial results on a novel technique for using SMT solvers to fine tune the weights of a ReLU neural network to guarantee outcomes on a finite set of particular examples. This procedure can be used to ensure performance on key examples, but it could also be used to insert difficult-to-find incorrect examples that trigger unexpected performance. We demonstrate this approach by fine tuning an MNIST network to incorrectly classify a particular image and discuss the potential for the approach to compromise reliability of freely-shared machine learning models.