Adversarial Robustness in Deep Learning: Attacks on Fragile Neurons

Pravin, C; Martino, I; Nicosia, G; Ojha, V

doi:10.1007/978-3-030-86362-3_2

Adversarial Robustness in Deep Learning: Attacks on Fragile Neurons

Lookup NU author(s): Dr Varun Ojha ORCiD

Downloads

Full text for this publication is not currently held within this repository. Alternative links are provided below where available.

Abstract

© 2021, Springer Nature Switzerland AG. We identify fragile and robust neurons of deep learning architectures using nodal dropouts of the first convolutional layer. Using an adversarial targeting algorithm, we correlate these neurons with the distribution of adversarial attacks on the network. Adversarial robustness of neural networks has gained significant attention in recent times and highlights an intrinsic weaknesses of deep learning networks against carefully constructed distortion applied to input images. In this paper, we evaluate the robustness of state-of-the-art image classification models trained on the MNIST and CIFAR10 datasets against the fast gradient sign method attack, a simple yet effective method of deceiving neural networks. Our method identifies the specific neurons of a network that are most affected by the adversarial attack being applied. We, therefore, propose to make fragile neurons more robust against these attacks by compressing features within robust neurons and amplifying the fragile neurons proportionally.