Confusing Deep Convolution Networks by Relabelling

TL;DR

This paper demonstrates how deep convolutional neural networks can be fooled into misclassifying images by subtle perturbations, effectively relabeling images without visual changes, exposing vulnerabilities in their generalization capabilities.

Contribution

It introduces a simple method to perturb images so they are visually indistinguishable but assigned arbitrary labels, revealing weaknesses in CNN robustness.

Findings

CNNs can be fooled into relabeling images with minimal perturbations

Perturbed images remain visually similar to originals

The method exposes vulnerabilities in CNN generalization

Abstract

Deep convolutional neural networks have become the gold standard for image recognition tasks, demonstrating many current state-of-the-art results and even achieving near-human level performance on some tasks. Despite this fact it has been shown that their strong generalisation qualities can be fooled to misclassify previously correctly classified natural images and give erroneous high confidence classifications to nonsense synthetic images. In this paper we extend that work, by presenting a straightforward way to perturb an image in such a way as to cause it to acquire any other label from within the dataset while leaving this perturbed image visually indistinguishable from the original.