Improving the Adversarial Robustness and Interpretability of Deep Neural Networks by Regularizing their Input Gradients

Keywords

Interpretability, Adversarial system, Robustness (evolution), Deep neural networks, Artificial intelligence

References

• Intriguing properties of neural networksChristian Szegedy, Wojciech Zaremba, Ilya Sutskever — 2013
• Decolorize: Fast, contrast enhancing, color to grayscale conversionMark Grundland, Neil A. Dodgson — 2007
• Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day ReadmissionRich Caruana, Yin Lou, Johannes Gehrke — 2015
• A Simple Method to Determine if a Music Information Retrieval System is a “Horse”Bob L. Sturm — 2014
• Improving generalization performance using double backpropagationHarris Drucker, Y. Le Cun — 1992
• Distillation as a Defense to Adversarial Perturbations Against Deep Neural NetworksNicolas Papernot, Patrick McDaniel, Xi Wu — 2015
• The Limitations of Deep Learning in Adversarial SettingsNicolas Papernot, Patrick McDaniel, Somesh Jha — 2015
• “Why Should I Trust You?”: Explaining the Predictions of Any ClassifierMarco Túlio Ribeiro, Sameer Singh, Carlos Guestrin — 2016
• Reading Digits in Natural Images with Unsupervised Feature LearningYuval Netzer — 2011
• Adversarial examples in the physical worldAlexey Kurakin, Ian Goodfellow, Samy Bengio — 2016
• Defensive Distillation is Not Robust to Adversarial ExamplesNicholas Carlini, David Wagner — 2016
• Cleverhans V0.1: an Adversarial Machine Learning LibraryIan Goodfellow, Nicolas Papernot, Patrick McDaniel — 2016
• Gradients of CounterfactualsMukund Sundararajan, Ankur Taly, Qiqi Yan — 2016
• Auditing black-box models for indirect influencePhilip Adler, Casey Falk, Sorelle A. Friedler — 2016
• Right for the Right Reasons: Training Differentiable Models by Constraining their ExplanationsAndrew Slavin Ross, Michael C. Hughes, Finale Doshi‐Velez — 2017
• Biologically inspired protection of deep networks from adversarial attacksAran Nayebi, Surya Ganguli — 2017
• Practical Black-Box Attacks against Machine LearningNicolas Papernot, Patrick McDaniel, Ian Goodfellow — 2016
• Feature Squeezing: Detecting Adversarial Examples in Deep Neural NetworksWeilin Xu, David Evans, Yanjun Qi — 2017
• Prediction of crime occurrence from multi-modal data using deep learningHyeon-Woo Kang, Hyeon-Woo Kang, Hang-Bong Kang — 2017
• The Space of Transferable Adversarial ExamplesFlorian Tramèr, Nicolas Papernot, Ian Goodfellow — 2017

Cited by

• Adversarial Detection of Flash Malware: Limitations and Open IssuesDavide Maiorca, Ambra Demontis, Battista Biggio — 2017
• Gradient Regularization Improves Accuracy of Discriminative ModelsDániel Varga, Adrián Csiszárik, Zsolt Zombori — 2017
• Threat of Adversarial Attacks on Deep Learning in Computer Vision: A SurveyNaveed Akhtar, Ajmal Mian — 2018
• Adversarial Vulnerability of Neural Networks Increases With Input DimensionCarl-Johann Simon-Gabriel, Yann Ollivier, Bernhard Schölkopf — 2018
• Understanding and Enhancing the Transferability of Adversarial ExamplesLei Wu, Zhanxing Zhu, Cheng Tai — 2018
• Improving DNN Robustness to Adversarial Attacks using Jacobian RegularizationDaniel Jakubovitz, Raja Giryes — 2018
• Security Consideration For Deep Learning-Based Image ForensicsWei Zhao, Pengpeng Yang, Rongrong Ni — 2018
• Towards Robust Training of Neural Networks by Regularizing Adversarial GradientsFuxun Yu, Zirui Xu, Yanzhi Wang — 2018
• Evaluating Feature Importance EstimatesSara Hooker, Dumitru Erhan, Pieter-Jan Kindermans — 2018
• Gradient Band-based Adversarial Training for Generalized Attack Immunity of A3C Path FindingTong Chen, Wenjia Niu, Yingxiao Xiang — 2018
• Defense Against Adversarial Attacks with Saak TransformSibo Song, Yueru Chen, Ngai‐Man Cheung — 2018
• Adversarial Examples: Attacks on Machine Learning-based Malware Visualization Detection MethodsXinbo Liu, Yapin Lin, Li He — 2018
• On the Intriguing Connections of Regularization, Input Gradients and Transferability of Evasion and Poisoning AttacksAmbra Demontis, Marco Melis, Maura Pintor — 2018
• Adversarial Examples: Opportunities and ChallengesJiliang Zhang, Chen Li — 2018
• Why the Failure? How Adversarial Examples Can Provide Insights for Interpretable Machine LearningRichard Tomsett, Amy Widdicombe, Tianwei Xing — 2018
• Interpreting Adversarial Robustness: A View from Decision Surface in Input SpaceFuxun Yu, Chenchen Liu, Yanzhi Wang — 2018
• Stakeholders in Explainable AIAlun Preece, Daniel Harborne, Dave Braines — 2018
• Training Machine Learning Models by Regularizing their ExplanationsAndrew Slavin Ross — 2018
• What can AI do for me?: evaluating machine learning interpretations in cooperative playShi Feng, Jordan Boyd‐Graber — 2018
• Efficient Two-Step Adversarial Defense for Deep Neural NetworksTing‐Jui Chang, Yukun He, Peng Li — 2018

Related papers

• Global Adversarial Attacks for Assessing Deep Learning RobustnessHanbin Hu, Mit Shah, Jianhua Z. Huang — 2019
• Developing and Defeating Adversarial ExamplesIan McDiarmid-Sterling, Allan Moser — 2020
• Adversarial Perturbation Defense on Deep Neural NetworksXingwei Zhang, Xiaolong Zheng, Wenji Mao — 2021
• Efficient Defenses Against Adversarial AttacksValentina Zantedeschi, Maria-Irina Nicolae, Ambrish Rawat — 2017
• Unifying Adversarial Training Algorithms with Flexible Deep Data Gradient RegularizationAlexander G. Ororbia, C. Lee Giles, Daniel Kifer — 2016