Consiglio Nazionale delle Ricerche

Tipo di prodottoArticolo in rivista
TitoloAdversarial image detection in deep neural networks
Anno di pubblicazione2019
FormatoElettronico
Autore/iCarrara F.; Falchi F.; Caldelli R.; Amato G.; Becarelli R.
Affiliazioni autoriCNR-ISTI, Pisa, Italy; CNR-ISTI, Pisa, Italy; University of Florence, Firenze, Italy; CNR-ISTI, Pisa, Italy; University of Florence, Firenze, Italy
Autori CNR e affiliazioni
  • FABIO CARRARA
  • GIUSEPPE AMATO
  • FABRIZIO FALCHI
Lingua/e
  • inglese
AbstractDeep neural networks are more and more pervading many computer vision applications and in particular image classification. Notwithstanding that, recent works have demonstrated that it is quite easy to create adversarial examples, i.e., images malevolently modified to cause deep neural networks to fail. Such images contain changes unnoticeable to the human eye but sufficient to mislead the network. This represents a serious threat for machine learning methods. In this paper, we investigate the robustness of the representations learned by the fooled neural network, analyzing the activations of its hidden layers. Specifically, we tested scoring approaches used for kNN classification, in order to distinguish between correctly classified authentic images and adversarial examples. These scores are obtained searching only between the very same images used for training the network. The results show that hidden layers activations can be used to reveal incorrect classifications caused by adversarial attacks.
Lingua abstractinglese
Altro abstract-
Lingua altro abstract-
Pagine da2815
Pagine a2835
Pagine totali21
RivistaMultimedia tools and applications
Attiva dal 1995
Editore: Kluwer Academic Publishers - Dordrecht ;
Paese di pubblicazione: Stati Uniti d'America
Lingua: inglese
ISSN: 1380-7501
Titolo chiave: Multimedia tools and applications
Numero volume della rivista78
Fascicolo della rivista3
DOI10.1007/s11042-018-5853-4
Verificato da refereeSì: Internazionale
Stato della pubblicazionePublished version
Indicizzazione (in banche dati controllate)
  • ISI Web of Science (WOS) (Codice:000458171600010)
  • Scopus (Codice:2-s2.0-85044183184)
Parole chiaveAdversarial images detection, Deep convolutional neural network, Machine learning security
Link (URL, URI)https://link.springer.com/article/10.1007%2Fs11042-018-5853-4
Titolo parallelo-
Licenza-
Scadenza embargo21 marzo 2019
Data di accettazione-
Note/Altre informazioni-
Strutture CNR
  • ISTI — Istituto di scienza e tecnologie dell'informazione "Alessandro Faedo"
Moduli/Attività/Sottoprogetti CNR-
Progetti Europei-
Allegati
Adversarial image detection in deep neural networks
Descrizione: Preprint
Tipo documento: application/pdf
Adversarial image detection in deep neural networks (documento privato )
Descrizione: published version
Tipo documento: application/pdf