Deep Networks with RBF Layers to Prevent Adversarial Examples

Authors: Petra Vidnerová, Roman Neruda
Publication year: 2018
Publication type: Konferenční příspěvek (mezinárodní konf.)
Periodical: Artificial Intelligence and Soft Computing
Projects: Capabilities and limitations of shallow and deep networks
Bibliographic citation: VIDNEROVÁ, Petra, NERUDA, Roman. Deep Networks with RBF Layers to Prevent Adversarial Examples. In: RUTKOWSKI, L., SCHERER, R., KORYTKOWSKI, M., PEDRYCZ, W., TADEUSIEWICZ, R., ZURADA, J. M., eds. Artificial Intelligence and Soft Computing. Cham: Springer, 2018, S. 257-266. ISBN 978-3-319-91252-3. ISSN 0302-9743. E-ISSN 1611-3349
Dostupné z: doi:10.1007/978-3-319-91253-0_25
DOI: 10.1007/978-3-319-91253-0_25
ID: 490841

Abstract

We propose a simple way to increase the robustness of deep neural network models to adversarial examples. The new architecture obtained by stacking deep neural network and RBF network is proposed. It is shown on experiments that such architecture is much more robust to adversarial examples than the original one while its accuracy on legitimate data stays more or less the same.