SINVAD: Search-based Image Space Navigation for DNN Image Classifier Test Input Generation

TL;DR

This paper introduces SINVAD, a search-based method that navigates the latent space of Variational Autoencoders to generate realistic test inputs for DNN image classifiers, improving robustness testing.

Contribution

It presents a novel approach using VAEs to explore plausible input spaces for more effective DNN testing beyond small perturbations.

Findings

Efficient generation of realistic test inputs.

Enhanced detection of DNN robustness issues.

Improved exploration of structured image space.

Abstract

The testing of Deep Neural Networks (DNNs) has become increasingly important as DNNs are widely adopted by safety critical systems. While many test adequacy criteria have been suggested, automated test input generation for many types of DNNs remains a challenge because the raw input space is too large to randomly sample or to navigate and search for plausible inputs. Consequently, current testing techniques for DNNs depend on small local perturbations to existing inputs, based on the metamorphic testing principle. We propose new ways to search not over the entire image space, but rather over a plausible input space that resembles the true training distribution. This space is constructed using Variational Autoencoders (VAEs), and navigated through their latent vector space. We show that this space helps efficiently produce test inputs that can reveal information about the robustness of DNNs when dealing with realistic tests, opening the field to meaningful exploration through the space of highly structured images.