東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Critical Pixels Attacks on Deep Neur...

Quan, Wei.

Linked to FindBook

Google Book

Amazon

博客來

Critical Pixels Attacks on Deep Neural Networks.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Critical Pixels Attacks on Deep Neural Networks./
Author:	Quan, Wei.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2020,
Description:	51 p.
Notes:	Source: Masters Abstracts International, Volume: 82-02.
Contained By:	Masters Abstracts International82-02.
Subject:	Computer engineering. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27964180
ISBN:	9798664702224

Critical Pixels Attacks on Deep Neural Networks.
Quan, Wei.

Critical Pixels Attacks on Deep Neural Networks. - Ann Arbor : ProQuest Dissertations & Theses, 2020 - 51 p.

Source: Masters Abstracts International, Volume: 82-02.

Thesis (M.S.)--State University of New York at Binghamton, 2020.

This item must not be sold to any third party vendors.

Deep neural networks (DNN) has been a hot research subject for years. It also has lots of practical applications. However, there are also a lot of concerns about the security problems and vulnerabilities of DNNs. Researchers have found that small perturbations can cause the DNN models to make mistakes. Many different methods have been discovered to produce those perturbations. In the field of image processing, when added to a clean image, these perturbations are not perceptible to human eyes but can fool a well-trained deep learning (DL) convolutional neural network (CNN) classifier. In this thesis, a new Critical-Pixel Iterative (CriPI) algorithm is introduced after a thorough study on the characteristics of critical pixels. The proposed CriPI algorithm is able to identify the critical pixels and generate one-pixel attack perturbations with a much higher efficiency. Comparing to the benchmark algorithm of one-pixel attack, the CriPI algorithm significantly reduces the time delay of the attack from seven minutes to one minute and achieves a similar performance.

ISBN: 9798664702224Subjects--Topical Terms:

621879
Computer engineering.
Subjects--Index Terms:

Adversarial attacks

Critical Pixels Attacks on Deep Neural Networks.
LDR:02150nmm a2200349 4500 001 2279705
005 20210823083423.5
008 220723s2020 ||||||||||||||||| ||eng d
020 $a 9798664702224
035 $a (MiAaPQ)AAI27964180
035 $a AAI27964180
040 $a MiAaPQ $c MiAaPQ
100 1 $a Quan, Wei. $3 2134112
245 1 0 $a Critical Pixels Attacks on Deep Neural Networks.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2020
300 $a 51 p.
500 $a Source: Masters Abstracts International, Volume: 82-02.
500 $a Advisor: Chen, Yu.
502 $a Thesis (M.S.)--State University of New York at Binghamton, 2020.
506 $a This item must not be sold to any third party vendors.
520 $a Deep neural networks (DNN) has been a hot research subject for years. It also has lots of practical applications. However, there are also a lot of concerns about the security problems and vulnerabilities of DNNs. Researchers have found that small perturbations can cause the DNN models to make mistakes. Many different methods have been discovered to produce those perturbations. In the field of image processing, when added to a clean image, these perturbations are not perceptible to human eyes but can fool a well-trained deep learning (DL) convolutional neural network (CNN) classifier. In this thesis, a new Critical-Pixel Iterative (CriPI) algorithm is introduced after a thorough study on the characteristics of critical pixels. The proposed CriPI algorithm is able to identify the critical pixels and generate one-pixel attack perturbations with a much higher efficiency. Comparing to the benchmark algorithm of one-pixel attack, the CriPI algorithm significantly reduces the time delay of the attack from seven minutes to one minute and achieves a similar performance.
590 $a School code: 0792.
650 4 $a Computer engineering. $3 621879
650 4 $a Artificial intelligence. $3 516317
653 $a Adversarial attacks
653 $a Critical Pixels
653 $a Deep Neural Networks
653 $a One pixel attack
690 $a 0464
690 $a 0800
710 2 $a State University of New York at Binghamton. $b Electrical and Computer Engineering. $3 3174435
773 0 $t Masters Abstracts International $g 82-02.
790 $a 0792
791 $a M.S.
792 $a 2020
793 $a English
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=27964180