Distribution Models for Falsification and Verification of DNNs
Authors: Felipe Toledo, David Shriver, Sebastian Elbaum, and Matthew B. DwyerAbstract:
DNN validation and verification approaches that are input distribution agnostic waste effort on irrelevant inputs and report false property violations. Drawing on the large body of work on model-based validation and verification of traditional systems, we introduce the first approach that leverages environmental models to focus DNN falsification and verification on the relevant input space. Our approach, DFV, automatically builds an input distribution model using unsupervised learning, prefixes that model to the DNN to force all inputs to come from the learned distribution, and reformulates the property to the input space of the distribution model. This transformed verification problem allows existing DNN falsification and verification tools to target the input distribution – avoiding consideration of infeasible inputs.
Recommended citation: Felipe Toledo, David Shriver, Sebastian Elbaum, Matthew B. Dwyer. 2021. Distribution Models for Falsification and Verification of DNNs. To Appear in ASE 2021