Probabilistic Jacobian-Based Saliency Maps Attacks - Archive ouverte HAL Access content directly
Journal Articles Machine Learning and Knowledge Extraction Year : 2020

Probabilistic Jacobian-Based Saliency Maps Attacks

Théo Combey
  • Function : Author
  • PersonId : 1086689
António Loison
  • Function : Author
  • PersonId : 1086690
Maxime Faucher
  • Function : Author
  • PersonId : 1086691
Hatem Hajri
  • Function : Author
  • PersonId : 1086606

Abstract

This paper introduces simple, faster and more efficient versions of the known targeted and untargeted Jacobian-based Saliency Map Attacks (JSMA). Despite creating adversarial examples with a higher average L 0 distance than the state-of-the-art Carlini-Wagner attack, the new versions of JSMA have a significant speed advantage over this attack, making them very convenient for L 0 real-time robustness testing of neural network classifiers.

Dates and versions

hal-03085884 , version 1 (22-12-2020)

Identifiers

Cite

Théo Combey, António Loison, Maxime Faucher, Hatem Hajri. Probabilistic Jacobian-Based Saliency Maps Attacks. Machine Learning and Knowledge Extraction, 2020, 2 (4), pp.558 - 578. ⟨10.3390/make2040030⟩. ⟨hal-03085884⟩
83 View
0 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More