Universal adversarial perturbations
This reveals potential security breaches in AI systems, as adversaries could exploit single input directions to break classifiers on most images, highlighting a foundational vulnerability in machine learning.
The paper tackled the problem of deep neural network classifiers being vulnerable to small, universal adversarial perturbations that cause misclassification of natural images with high probability, showing that such perturbations exist and generalize well across networks.
Given a state-of-the-art deep neural network classifier, we show the existence of a universal (image-agnostic) and very small perturbation vector that causes natural images to be misclassified with high probability. We propose a systematic algorithm for computing universal perturbations, and show that state-of-the-art deep neural networks are highly vulnerable to such perturbations, albeit being quasi-imperceptible to the human eye. We further empirically analyze these universal perturbations and show, in particular, that they generalize very well across neural networks. The surprising existence of universal perturbations reveals important geometric correlations among the high-dimensional decision boundary of classifiers. It further outlines potential security breaches with the existence of single directions in the input space that adversaries can possibly exploit to break a classifier on most natural images.