Learning Universal Adversarial Perturbations with Generative Models
Authors:
- Jamie HayesJHJamie Hayes
- George DanezisGDGeorge Danezis
- 1 May 2018
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 43-49
Abstract
Neural networks are known to be vulnerable to adversarial examples, inputs that have been intentionally perturbed to remain visually similar to the source input, but cause a misclassification. It was recently shown that given a dataset and classifier, there exists so called universal adversarial perturbations, a single perturbation that causes a misclassification when applied to any input. In this work, we introduce universal adversarial networks, a generative network that is capable of fooling a target classifier when it's generated output is added to a clean sample from a dataset. We show that this technique improves on known universal adversarial attacks.Keywords
All Related Versions
This publication has 19 references indexed in Scilit:
- Densely Connected Convolutional Networks2017
- Rethinking the Inception Architecture for Computer Vision2016
- Adversarial Feature Selection Against Evasion AttacksIEEE Transactions on Cybernetics, 2015
- A rapid learning algorithm for vehicle classificationInformation Sciences, 2015
- Active learning for on-road vehicle detection: a comparative studyMachine Vision and Applications, 2011
- Machine Learning in Financial Crisis Prediction: A SurveyIEEE Transactions on Systems, Man and Cybernetics, Part C (Applications and Reviews), 2011
- Mean squared error: Love it or leave it? A new look at Signal Fidelity MeasuresIEEE Signal Processing Magazine, 2009
- A Minimax Theorem with Applications to Machine Learning, Signal Processing, and FinanceSIAM Journal on Optimization, 2008
- Predicting hepatitis B virus–positive metastatic hepatocellular carcinomas using gene expression profiling and supervised machine learningNature Medicine, 2003
- Support vector machine for regression and applications to financial forecasting2000