On the Robustness of the CVPR 2018 White-Box Adversarial Example Defenses

Neural networks are known to be vulnerable to adversarial examples. In this note, we evaluate the two white-box defenses that appeared at CVPR 2018 and find they are ineffective: when applying existing techniques, we can reduce the accuracy of the defended models to 0%.

AI hardware & compute Security & robustness cs.CV cs.CR cs.LG stat.ML Adversarial Robustness in Machine Learning Physical Unclonable Functions (PUFs) and Hardware Security Cryptographic Implementations and Security Computer Science Mathematics

View graph Source

Paper

References (8)

01Adversarial Risk and the Dangers of Evaluating Against Weak AttacksJonathan Uesato, Brendan O'Donoghue, Aäron van den Oord et al.2018 · International Conference on Machine Learning · 663 citations In Library

02Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial ExamplesAnish Athalye, Nicholas Carlini, D. Wagner2018 · International Conference on Machine Learning · 3.6k citations In Library

03Deflecting Adversarial Attacks with Pixel DeflectionAaditya (Adi) Prakash, N. Moran, Solomon Garber et al.2018 · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · 334 citations In Library

04Defense Against Adversarial Attacks Using High-Level Representation Guided DenoiserFangzhou Liao, Ming Liang, Yinpeng Dong et al.2017 · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · 1.0k citations In Library

05Towards Deep Learning Models Resistant to Adversarial AttacksA. Ma̧dry, Aleksandar Makelov, Ludwig Schmidt et al.2017 · International Conference on Learning Representations · 15k citations In Library

06Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection MethodsNicholas Carlini, D. Wagner2017 · AISec@CCS · 2.0k citations In Library

07Intriguing properties of neural networksChristian Szegedy, Wojciech Zaremba, I. Sutskever et al.2013 · International Conference on Learning Representations · 17k citations In Library

08ImageNet: A large-scale hierarchical image databaseJia Deng, Wei Dong, R. Socher et al.2009 · 2009 IEEE Conference on Computer Vision and Pattern Recognition · 75k citations

Similar papers

Beware the Black-Box: On the Robustness of Recent Defenses to Adversarial ExamplesKaleel Mahmood, Deniz Gurevin, Marten van Dijk et al.2021 · Entropy In Library

Beware the Black-Box: on the Robustness of Recent Defenses to Adversarial ExamplesKaleel Mahmood2021 · arXiv (Cornell University) · 30 citations In Library

On the Effectiveness of Adversarial Training in Defending against Adversarial Example Attacks for Image ClassificationSanglee Park, Jungmin So2020 · Applied Sciences · 28 citations In Library

Impact of White-Box Adversarial Attacks on Convolutional Neural NetworksRakesh Podder, Sudipto Ghosh2024 · arXiv (Cornell University) · 2 citations In Library

Impact of White-Box Adversarial Attacks on Convolutional Neural NetworksRakesh Podder, Sudipto Ghosh2024 In Library

An Empirical Evaluation of Adversarial Examples Defences, Combinations and Robustness ScoresAleksandar Janković, Rudolf Mayer2022 · IWSPA@CODASPY · 3 citations In Library

The Vulnerability of the Neural Networks Against Adversarial Examples in Deep Learning AlgorithmsRui Zhao2021 · 2021 2nd International Conference on Computing and Data Science (CDS) In Library

The Vulnerability of the Neural Networks Against Adversarial Examples in Deep Learning AlgorithmsRui Zhao2020 · arXiv (Cornell University) · 1 citation In Library

A Brief Comparison Between White Box, Targeted Adversarial Attacks in Deep Neural NetworksGrigor Bezirganyan, Henrik Sergoyan2022 · Mathematical Problems of Computer Science In Library

Directed Adversarial Training for Comprehensive Robustification of CNNsBhuvi Ghosh, Harsh Shetye, Atharv Salian et al.2024 · International Conference on Computing Communication and Networking Technologies · 0 citations In Library

Source of record

arXiv

44B

Welcome.

This panel sticks with you. Pick anything below and it opens right beside it, so you can dig through 60,000-plus records without ever losing your spot here.

Travel 44B

LibraryPapers, policy, standards, statute LabsEvery organization building AI ModelsIntelligence, price, and speed BenchmarksThe evaluation catalog IncidentsAI failures and harms SearchOne field across everything DashboardThe state of AI accountability in NY 44B RegistryThe Art. 44-B compliance portal