SSCNets: Robustifying DNNs using Secure Selective Convolutional Filters

被引：11

作者：

Ali, Hassan ^{[1
]}

Khalid, Faiq ^{[2
]}

Tariq, Hammad Ali ^{[1
]}

Hanif, Muhammad Abdullah ^{[2
]}

Ahmed, Rehan ^{[1
]}

Rehman, Semeen ^{[2
]}

机构：

[1] Natl Univ Sci & Technol NUST, Islamabad, Pakistan

[2] Tech Univ Wien TU Wien, Vienna, Austria

来源：

IEEE DESIGN & TEST | 2020年 / 37卷 / 02期

关键词：

Robustness; Image edge detection; Training data; Deep learning; Perturbation methods; Filtering; Feature extraction; Convolutional Neural Network; CNN; Sobel Filters; Adversarial Attacks; Defenses; Machine Learning; ML Security; high-pass filters;

D O I：

10.1109/MDAT.2019.2961325

中图分类号：

TP3 [计算技术、计算机技术];

学科分类号：

0812 ;

摘要：

Training data is crucial in ensuring robust neural inference, and deep neural networks (DNNs) are heavily dependent on this assumption. However, DNNs can be exploited by adversaries that facilitate various attacks. Adversarial defenses include several techniques, some of which happen during the preprocessing stages (i.e., noise filtering, etc.). This article analyzes the impact of some preprocessing filters, and proposes a selective preprocessing method which increases robustness and reduces the computational complexity.-Theocharis Theocharides, University of Cyprus

引用

页码：58 / 65

页数：8

共 12 条

[1]

[Anonymous], ARXIV170704131

[2]

Athalye A, 2018, PR MACH LEARN RES, V80

[3]

Brendel W., 2018, PROC 6 INT C LEARN R

[4] Towards Evaluating the Robustness of Neural Networks [J].

Carlini, Nicholas ;

Wagner, David .

2017 IEEE SYMPOSIUM ON SECURITY AND PRIVACY (SP), 2017, :39-57

[5]

Cheng M., 2018, Query-efficient hard-label black-box attack:an optimization-based approach

[6]

Ghazikhani H, 2018, 2018 8TH INTERNATIONAL CONFERENCE ON COMPUTER AND KNOWLEDGE ENGINEERING (ICCKE), P1, DOI 10.1109/ICCKE.2018.8566534

[7]

Goodfellow I. J., 2015, P INT C LEARN REPR

[8]

Khalid F, 2019, DES AUT TEST EUROPE, P902, DOI [10.23919/date.2019.8715141, 10.23919/DATE.2019.8715141]

[9]

Kurach Karol, 2019, PR MACH LEARN RES, V97

[10] The Limitations of Deep Learning in Adversarial Settings [J].

Papernot, Nicolas ;

McDaniel, Patrick ;

Jha, Somesh ;

Fredrikson, Matt ;

Celik, Z. Berkay ;

Swami, Ananthram .

1ST IEEE EUROPEAN SYMPOSIUM ON SECURITY AND PRIVACY, 2016, :372-387

← 1 2 →