An Adversarial Robustness Perspective on the Topology of Neural Networks - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

An Adversarial Robustness Perspective on the Topology of Neural Networks

Résumé

In this paper, we investigate the impact of neural networks (NNs) topology on adversarial robustness. Specifically, we study the graph produced when an input traverses all the layers of a NN, and show that such graphs are different for clean and adversarial inputs. We find that graphs from clean inputs are more centralized around highway edges, whereas those from adversaries are more diffuse, leveraging under-optimized edges. Through experiments on a variety of datasets and architectures, we show that these under-optimized edges are a source of adversarial vulnerability and that they can be used to detect adversarial inputs.

Dates et versions

hal-03959759 , version 1 (27-01-2023)

Identifiants

Citer

Morgane Goibert, Thomas Ricatte, Elvis Dohmatob. An Adversarial Robustness Perspective on the Topology of Neural Networks. ML Safety Workshop, NeurIPS 2022, Dec 2022, New Orleans, United States. ⟨hal-03959759⟩
12 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More