Practical black-box attacks on deep neural networks using efficient query mechanisms

Arjun Nitin Bhagoji, Warren He, Bo Li, Dawn Song

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Existing black-box attacks on deep neural networks (DNNs) have largely focused on transferability, where an adversarial instance generated for a locally trained model can “transfer” to attack other learning models. In this paper, we propose novel Gradient Estimation black-box attacks for adversaries with query access to the target model’s class probabilities, which do not rely on transferability. We also propose strategies to decouple the number of queries required to generate each adversarial sample from the dimensionality of the input. An iterative variant of our attack achieves close to 100% attack success rates for both targeted and untargeted attacks on DNNs. We carry out a thorough comparative evaluation of black-box attacks and show that Gradient Estimation attacks achieve attack success rates similar to state-of-the-art white-box attacks on the MNIST and CIFAR-10 datasets. We also apply the Gradient Estimation attacks successfully against real-world classifiers hosted by Clarifai. Further, we evaluate black-box attacks against state-of-the-art defenses based on adversarial training and show that the Gradient Estimation attacks are very effective even against these defenses.

Original languageEnglish (US)
Title of host publicationComputer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
EditorsMartial Hebert, Vittorio Ferrari, Cristian Sminchisescu, Yair Weiss
PublisherSpringer-Verlag Berlin Heidelberg
Pages158-174
Number of pages17
ISBN (Print)9783030012571
DOIs
StatePublished - 2018
Event15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany
Duration: Sep 8 2018Sep 14 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11216 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other15th European Conference on Computer Vision, ECCV 2018
CountryGermany
CityMunich
Period9/8/189/14/18

Keywords

  • Adversarial examples
  • Black-box attacks
  • Deep neural networks
  • Image classification

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Practical black-box attacks on deep neural networks using efficient query mechanisms'. Together they form a unique fingerprint.

Cite this