prikaz prve stranice dokumenta Obrane klasifikacijskih modela od napada trovanjem podataka
No public access
undergraduate thesis
Obrane klasifikacijskih modela od napada trovanjem podataka
Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing, 2023. urn:nbn:hr:168:445913
Capan, Nikola
University of Zagreb
Faculty of Electrical Engineering and Computing

Institutional repository: FER Repository

Cite this document

Capan, N. (2023). Obrane klasifikacijskih modela od napada trovanjem podataka (Undergraduate thesis). Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing. Retrieved from https://urn.nsk.hr/urn:nbn:hr:168:445913

Capan, Nikola. "Obrane klasifikacijskih modela od napada trovanjem podataka." Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2023. https://urn.nsk.hr/urn:nbn:hr:168:445913

Capan, Nikola. "Obrane klasifikacijskih modela od napada trovanjem podataka." Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2023. https://urn.nsk.hr/urn:nbn:hr:168:445913

Capan, N. (2023). 'Obrane klasifikacijskih modela od napada trovanjem podataka', Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, accessed 15 August 2024, https://urn.nsk.hr/urn:nbn:hr:168:445913

Capan N. Obrane klasifikacijskih modela od napada trovanjem podataka [Undergraduate thesis]. Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing; 2023 [cited 2024 August 15] Available at: https://urn.nsk.hr/urn:nbn:hr:168:445913

N. Capan, "Obrane klasifikacijskih modela od napada trovanjem podataka", Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, 2023. Available at: https://urn.nsk.hr/urn:nbn:hr:168:445913

Metadata
TitleObrane klasifikacijskih modela od napada trovanjem podataka
Title (english)Defence of classification models from data-poisoning attacks
AuthorNikola Capan
MentorSiniša Šegvić (mentor)
Committee memberSiniša Šegvić (član povjerenstva)
Committee memberIvan Grubišić (član povjerenstva)
Committee memberZorana Doždor (član povjerenstva)
GranterUniversity of Zagreb
Faculty of Electrical Engineering and Computing
Zagreb
Defense date and country2023-07-06, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
Ovaj rad se bavi napadima na duboke konvolucijske modele trovanjem podataka za treniranje. Cilj takvih napada je istrenirati model da u normalnim okolnostima radi dobro i ispravno, ali u prisustvu određenog okidača, radi u skladu s napadačevim ciljevima. Dosadašnji takvi napadi su uglavnom taj okidač dizajnirali tako da je on ograničen na mali prostor slike. U ovom radu, okidač koji je korišten se proteže po cijeloj slici i izuzetno je neprimjetan ljudskom oku. To je izvedeno modifikacijom piksela u frekvencijskoj domeni slika, umjesto u prostornoj. Ovakav napad se pokazao izuzetno uspješnim na skupovima MNIST, CIFAR10 i PubFig. Mehanizam obrane koji je uspio drastično umanjiti uspješnost napada je jednostavno filtriranje svih slika za treniranje Gaussovim filtrom. Model naučen na takvom skupu, s obzirom da su slike puno mutnije, pokazuje nešto lošije rezultate i na čistim, neotrovanim slikama.
Abstract (english)
This paper deals with the attacks on deep convolutional models by poisoning the training dana. The goal of such attacks is to train a model to behave normally in usual conditions, but in presence of a particular trigger, do what the attacker wants. The triggers in existing such attacks were mostly spatially limited on training images. Trigger that was used in this paper is dispersed across whole image which makes it invisable to humans. It was accomplished by modifying the pixels in frequency domain of training images, instead of spatial. This attack has proven to be successful on MNIST, CIFAR and PubFig training sets. Defence mechanism that significantly decreased the success of the attack is simple filtering all of the training images with Gaussian filter. The model trained on such filtered set shows slightly worse results on clean, unpoisoned images, because the training images were blurrier.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:168:445913
Study programmeTitle: Computing
Study programme type: university
Study level: undergraduate
Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsClosed access
Terms of useLicence In copyright icon
Public note
Created on2023-10-11 08:35:13