000268674 001__ 268674
000268674 005__ 20190812204806.0
000268674 02470 $$2ArXiv$$a1903.10826
000268674 037__ $$aCONF
000268674 245__ $$aA geometry-inspired decision-based attack
000268674 260__ $$c2019
000268674 269__ $$a2019
000268674 300__ $$a10
000268674 336__ $$aConference Papers
000268674 520__ $$aDeep neural networks have recently achieved tremen-dous success in image classification. Recent studies havehowever shown that they are easily misled into incorrectclassification decisions by adversarial examples. Adver-saries can even craft attacks by querying the model in black-box settings, where no information about the model is re-leased except its final decision. Such decision-based at-tacks usually require lots of queries, while real-world imagerecognition systems might actually restrict the number ofqueries. In this paper, we propose qFool, a novel decision-based attack algorithm that can generate adversarial exam-ples using a small number of queries. The qFool method candrastically reduce the number of queries compared to pre-vious decision-based attacks while reaching the same qual-ity of adversarial examples. We also enhance our methodby constraining adversarial perturbations in low-frequencysubspace, which can make qFool even more computation-ally efficient. Altogether, we manage to fool commercialimage recognition systems with a small number of queries,which demonstrates the actual effectiveness of our new al-gorithm in practice.
000268674 6531_ $$aml-tm
000268674 700__ $$aLiu, Yujia
000268674 700__ $$aMoosavi Dezfooli, Seyed Mohsen$$0249003$$g226282
000268674 700__ $$0241061$$aFrossard, Pascal$$g101475
000268674 7112_ $$dOct 27, 2019 - Nov 3, 2019$$cSeoul, South Korea$$aICCV 2019 : IEEE International Conference on Computer Vision
000268674 773__ $$t[Proceedings of ICCV 2019]
000268674 8560_ $$fpascal.frossard@epfl.ch
000268674 909C0 $$pLTS4$$mpascal.frossard@epfl.ch$$0252393$$zMarselli, Béatrice$$xU10851
000268674 909CO $$pconf$$pSTI$$ooai:infoscience.epfl.ch:268674
000268674 960__ $$apascal.frossard@epfl.ch
000268674 961__ $$aalessandra.bianchi@epfl.ch
000268674 973__ $$aEPFL$$rREVIEWED
000268674 980__ $$aCONF
000268674 981__ $$aoverwrite