Comparison of deep neural network fooling methods on the accuracy of classification

Witold Oleszkiewicz

doi:10.1117/12.2501586

1 October 2018 Comparison of deep neural network fooling methods on the accuracy of classification

Witold Oleszkiewicz

Proceedings Volume 10808, Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 2018; 108082I (2018) https://doi.org/10.1117/12.2501586
Event: Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 2018, 2018, Wilga, Poland

Abstract

The ability to train neural networks depends on access to data. In some areas, for example in medicine, it is difficult to obtain large datasets since medical data can contain very sensitive information. It is desirable to anonymize the dataset in such a way that the utility of machine learning prediction models is preserved. In this paper, we compare different methods of fooling deep neural networks. We investigate how different algorithms affects the accuracy of one classification task while fooling classifier in the other classification task.

Citation Download Citation

Witold Oleszkiewicz "Comparison of deep neural network fooling methods on the accuracy of classification", Proc. SPIE 10808, Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 2018, 108082I (1 October 2018); https://doi.org/10.1117/12.2501586

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available