Generating Minimalist Adversarial Perturbations to Test Object-Detection Models: An Adaptive Multi-Metric Evolutionary Search Approach

TL;DR

This paper presents TM-EVO, an evolutionary algorithm that efficiently generates minimal adversarial perturbations to test object-detection models, revealing their vulnerabilities with less noise than existing methods.

Contribution

Introduction of TM-EVO, a multi-metric evolutionary search method for creating minimal adversarial perturbations to evaluate object detection models' robustness.

Findings

TM-EVO outperforms baseline in generating less noisy adversarial examples.

TM-EVO is efficient in evaluating models like DETR and Faster R-CNN.

Adversarial tests reveal vulnerabilities in popular object detection models.

Abstract

Deep Learning (DL) models excel in computer vision tasks but can be susceptible to adversarial examples. This paper introduces Triple-Metric EvoAttack (TM-EVO), an efficient algorithm for evaluating the robustness of object-detection DL models against adversarial attacks. TM-EVO utilizes a multi-metric fitness function to guide an evolutionary search efficiently in creating effective adversarial test inputs with minimal perturbations. We evaluate TM-EVO on widely-used object-detection DL models, DETR and Faster R-CNN, and open-source datasets, COCO and KITTI. Our findings reveal that TM-EVO outperforms the state-of-the-art EvoAttack baseline, leading to adversarial tests with less noise while maintaining efficiency.