Noise Based Approach for the Detection of Adversarial Examples

We propose a new method for detecting adversarial examples based on a stochastic approach. An example is presented to the network several times and classified as adversarial if the fraction of times the output label is different from the label generated by the deterministic network is above some thr...

Descripción completa

Detalles Bibliográficos
Autores principales: Kloster, Matias Alejandro, Cúñale, Ariel Hernán, Mato, Germán
Formato: Objeto de conferencia
Lenguaje:Inglés
Publicado: 2020
Materias:
Ciencias Informáticas
Adversarial examples
Method for detecting
Acceso en línea:http://sedici.unlp.edu.ar/handle/10915/116415
http://49jaiio.sadio.org.ar/pdfs/agranda/AGRANDA-04.pdf
Aporte de:
SEDICI (UNLP) de Universidad Nacional de La Plata
id I19-R120-10915-116415
record_format dspace
institution Universidad Nacional de La Plata
institution_str I-19
repository_str R-120
collection SEDICI (UNLP)
language Inglés
topic Ciencias Informáticas
Adversarial examples
Method for detecting
spellingShingle Ciencias Informáticas
Adversarial examples
Method for detecting
Kloster, Matias Alejandro
Cúñale, Ariel Hernán
Mato, Germán
Noise Based Approach for the Detection of Adversarial Examples
topic_facet Ciencias Informáticas
Adversarial examples
Method for detecting
description We propose a new method for detecting adversarial examples based on a stochastic approach. An example is presented to the network several times and classified as adversarial if the fraction of times the output label is different from the label generated by the deterministic network is above some threshold value. We analyze the performance of the method for three attack methods (DeepFool, Fast Gradient Sign Method and norm 2 Carlini Wagner) and two datasets (MNIST and CIFAR-10). We find that our approach works best for stronger attacks such as DeepFool and CW2, and could be used as part of a scheme where several methods are applied simultaneously in order to estimate if a given input is adversarial or not.
format Objeto de conferencia
Objeto de conferencia
author Kloster, Matias Alejandro
Cúñale, Ariel Hernán
Mato, Germán
author_facet Kloster, Matias Alejandro
Cúñale, Ariel Hernán
Mato, Germán
author_sort Kloster, Matias Alejandro
title Noise Based Approach for the Detection of Adversarial Examples
title_short Noise Based Approach for the Detection of Adversarial Examples
title_full Noise Based Approach for the Detection of Adversarial Examples
title_fullStr Noise Based Approach for the Detection of Adversarial Examples
title_full_unstemmed Noise Based Approach for the Detection of Adversarial Examples
title_sort noise based approach for the detection of adversarial examples
publishDate 2020
url http://sedici.unlp.edu.ar/handle/10915/116415
http://49jaiio.sadio.org.ar/pdfs/agranda/AGRANDA-04.pdf
work_keys_str_mv AT klostermatiasalejandro noisebasedapproachforthedetectionofadversarialexamples
AT cunalearielhernan noisebasedapproachforthedetectionofadversarialexamples
AT matogerman noisebasedapproachforthedetectionofadversarialexamples
bdutipo_str Repositorios
_version_ 1764820446990041089

Ejemplares similares