The race to robustness: exploiting fragile models for urban camouflage and the imperative for machine learning security

TL;DR

This paper introduces Distributed Adversarial Regions (DAR), a simple yet effective AML attack method that highlights the fragility of urban object detection models and underscores the need for improved ML security.

Contribution

It presents a novel distributed adversarial attack technique and compares it with existing AML methods in urban camouflage scenarios.

Findings

DAR reduces model confidence by 40.4% on average.

DAR does not require perturbing the entire image or focal object.

The method demonstrates how easily models can be attacked with minimal skill.

Abstract

Adversarial Machine Learning (AML) represents the ability to disrupt Machine Learning (ML) algorithms through a range of methods that broadly exploit the architecture of deep learning optimisation. This paper presents Distributed Adversarial Regions (DAR), a novel method that implements distributed instantiations of computer vision-based AML attack methods that may be used to disguise objects from image recognition in both white and black box settings. We consider the context of object detection models used in urban environments, and benchmark the MobileNetV2, NasNetMobile and DenseNet169 models against a subset of relevant images from the ImageNet dataset. We evaluate optimal parameters (size, number and perturbation method), and compare to state-of-the-art AML techniques that perturb the entire image. We find that DARs can cause a reduction in confidence of 40.4% on average, but with the benefit of not requiring the entire image, or the focal object, to be perturbed. The DAR method is a deliberately simple approach where the intention is to highlight how an adversary with very little skill could attack models that may already be productionised, and to emphasise the fragility of foundational object detection models. We present this as a contribution to the field of ML security as well as AML. This paper contributes a novel adversarial method, an original comparison between DARs and other AML methods, and frames it in a new context - that of urban camouflage and the necessity for ML security and model robustness.