You are here

UCF Digital Collections ยป ETD Collection

Adversarial Attacks On Vision Algorithms Using Deep Learning Features

Download pdf | Full Screen View

Date Issued:
2017
Abstract/Description:
Computer vision algorithms, such as those implementing object detection, are known to be sus-ceptible to adversarial attacks. Small barely perceptible perturbations to the input can cause visionalgorithms to incorrectly classify inputs that they would have otherwise classified correctly. Anumber of approaches have been recently investigated to generate such adversarial examples fordeep neural networks. Many of these approaches either require grey-box access to the deep neuralnet being attacked or rely on adversarial transfer and grey-box access to a surrogate neural network.In this thesis, we present an approach to the synthesis of adversarial examples for computer vi-sion algorithms that only requires black-box access to the algorithm being attacked. Our attackapproach employs fuzzing with features derived from the layers of a convolutional neural networktrained on adversarial examples from an unrelated dataset. Based on our experimental results,we believe that our validation approach will enable designers of cyber-physical systems and otherhigh-assurance use-cases of vision algorithms to stress test their implementations.
Title: Adversarial Attacks On Vision Algorithms Using Deep Learning Features.
89 views
19 downloads
Name(s): Michel, Andy, Author
Jha, Sumit Kumar, Committee Chair
Leavens, Gary, Committee Member
Valliyil Thankachan, Sharma, Committee Member
University of Central Florida, Degree Grantor
Type of Resource: text
Date Issued: 2017
Publisher: University of Central Florida
Language(s): English
Abstract/Description: Computer vision algorithms, such as those implementing object detection, are known to be sus-ceptible to adversarial attacks. Small barely perceptible perturbations to the input can cause visionalgorithms to incorrectly classify inputs that they would have otherwise classified correctly. Anumber of approaches have been recently investigated to generate such adversarial examples fordeep neural networks. Many of these approaches either require grey-box access to the deep neuralnet being attacked or rely on adversarial transfer and grey-box access to a surrogate neural network.In this thesis, we present an approach to the synthesis of adversarial examples for computer vi-sion algorithms that only requires black-box access to the algorithm being attacked. Our attackapproach employs fuzzing with features derived from the layers of a convolutional neural networktrained on adversarial examples from an unrelated dataset. Based on our experimental results,we believe that our validation approach will enable designers of cyber-physical systems and otherhigh-assurance use-cases of vision algorithms to stress test their implementations.
Identifier: CFE0006898 (IID), ucf:51714 (fedora)
Note(s): 2017-12-01
M.S.
Engineering and Computer Science, Computer Science
Masters
This record was generated from author submitted information.
Subject(s): Deep Learning -- Computer Vision -- Adversarial Attack
Persistent Link to This Record: http://purl.flvc.org/ucf/fd/CFE0006898
Restrictions on Access: public 2017-12-15
Host Institution: UCF

In Collections