Deep neural networks are easily fooled: High confidence predictions for unrecognizable images

TLDR

Deep neural networks now achieve near‑human accuracy on visual classification tasks, prompting questions about how computer vision differs from human perception. The study demonstrates that it is trivial to generate images that are unrecognizable to humans yet are classified by state‑of‑the‑art DNNs with 99.99 % confidence. Using convolutional neural networks trained on ImageNet or MNIST, the authors employ evolutionary algorithms or gradient ascent to produce images that the networks assign high‑confidence labels to. These fooling images reveal that DNNs can confidently misclassify unrecognizable inputs, underscoring fundamental differences between human and machine vision and questioning the generality of current computer‑vision models.

Abstract

Deep neural networks (DNNs) have recently been achieving state-of-the-art performance on a variety of pattern-recognition tasks, most notably visual classification problems. Given that DNNs are now able to classify objects in images with near-human-level performance, questions naturally arise as to what differences remain between computer and human vision. A recent study [30] revealed that changing an image (e.g. of a lion) in a way imperceptible to humans can cause a DNN to label the image as something else entirely (e.g. mislabeling a lion). Here we show a related result: it is easy to produce images that are completely unrecognizable to humans, but that state-of-the-art DNNs believe to be recognizable objects with 99.99% confidence (e.g. labeling with certainty that white noise static is a lion). Specifically, we take convolutional neural networks trained to perform well on either the ImageNet or MNIST datasets and then find images with evolutionary algorithms or gradient ascent that DNNs label with high confidence as belonging to each dataset class. It is possible to produce images totally unrecognizable to human eyes that DNNs believe with near certainty are familiar objects, which we call “fooling images” (more generally, fooling examples). Our results shed light on interesting differences between human vision and current DNNs, and raise questions about the generality of DNN computer vision.