An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines

Wen Su; Qingna Li

doi:10.1109/DSIT55514.2022.9943927

An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines

Wen Su, Qingna Li^*

^*Corresponding author for this work

School of Mathematics and Statistics

Beijing Institute of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Adversarial perturbations have drawn great attentions in various machine learning models. In this paper, we investigate the sample adversarial perturbations for nonlinear support vector machines (SVMs). Due to the implicit form of the nonlinear functions mapping data to the feature space, it is difficult to obtain the explicit form of the adversarial perturbations. By exploring the special property of nonlinear SVMs, we transform the optimization problem of attacking nonlinear SVMs into a nonlinear KKT system. Such a system can be solved by various numerical methods. Numerical results show that our method is efficient in computing adversarial perturbations.

Original language	English
Title of host publication	2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781665498685
DOIs	https://doi.org/10.1109/DSIT55514.2022.9943927
Publication status	Published - 2022
Event	5th International Conference on Data Science and Information Technology, DSIT 2022 - Shanghai, China Duration: 22 Jul 2022 → 24 Jul 2022

Publication series

Name	2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings

Conference

Conference	5th International Conference on Data Science and Information Technology, DSIT 2022
Country/Territory	China
City	Shanghai
Period	22/07/22 → 24/07/22

Keywords

KKT system
adversarial perturbation
machine learning
nonlinear optimization
support vector machine

Access to Document

10.1109/DSIT55514.2022.9943927

Cite this

Su, W., & Li, Q. (2022). An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines. In 2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings (2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/DSIT55514.2022.9943927

Su, Wen ; Li, Qingna. / An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines. 2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2022. (2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings).

@inproceedings{e3abd83a53d14d229882df95abe1ab59,

title = "An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines",

abstract = "Adversarial perturbations have drawn great attentions in various machine learning models. In this paper, we investigate the sample adversarial perturbations for nonlinear support vector machines (SVMs). Due to the implicit form of the nonlinear functions mapping data to the feature space, it is difficult to obtain the explicit form of the adversarial perturbations. By exploring the special property of nonlinear SVMs, we transform the optimization problem of attacking nonlinear SVMs into a nonlinear KKT system. Such a system can be solved by various numerical methods. Numerical results show that our method is efficient in computing adversarial perturbations.",

keywords = "KKT system, adversarial perturbation, machine learning, nonlinear optimization, support vector machine",

author = "Wen Su and Qingna Li",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 5th International Conference on Data Science and Information Technology, DSIT 2022 ; Conference date: 22-07-2022 Through 24-07-2022",

year = "2022",

doi = "10.1109/DSIT55514.2022.9943927",

language = "English",

series = "2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings",

address = "United States",

}

Su, W & Li, Q 2022, An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines. in 2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings. 2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings, Institute of Electrical and Electronics Engineers Inc., 5th International Conference on Data Science and Information Technology, DSIT 2022, Shanghai, China, 22/07/22. https://doi.org/10.1109/DSIT55514.2022.9943927

An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines. / Su, Wen; Li, Qingna.
2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2022. (2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines

AU - Su, Wen

AU - Li, Qingna

PY - 2022

Y1 - 2022

N2 - Adversarial perturbations have drawn great attentions in various machine learning models. In this paper, we investigate the sample adversarial perturbations for nonlinear support vector machines (SVMs). Due to the implicit form of the nonlinear functions mapping data to the feature space, it is difficult to obtain the explicit form of the adversarial perturbations. By exploring the special property of nonlinear SVMs, we transform the optimization problem of attacking nonlinear SVMs into a nonlinear KKT system. Such a system can be solved by various numerical methods. Numerical results show that our method is efficient in computing adversarial perturbations.

AB - Adversarial perturbations have drawn great attentions in various machine learning models. In this paper, we investigate the sample adversarial perturbations for nonlinear support vector machines (SVMs). Due to the implicit form of the nonlinear functions mapping data to the feature space, it is difficult to obtain the explicit form of the adversarial perturbations. By exploring the special property of nonlinear SVMs, we transform the optimization problem of attacking nonlinear SVMs into a nonlinear KKT system. Such a system can be solved by various numerical methods. Numerical results show that our method is efficient in computing adversarial perturbations.

KW - KKT system

KW - adversarial perturbation

KW - machine learning

KW - nonlinear optimization

KW - support vector machine

UR - http://www.scopus.com/inward/record.url?scp=85143124752&partnerID=8YFLogxK

U2 - 10.1109/DSIT55514.2022.9943927

DO - 10.1109/DSIT55514.2022.9943927

M3 - Conference contribution

AN - SCOPUS:85143124752

T3 - 2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings

BT - 2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 5th International Conference on Data Science and Information Technology, DSIT 2022

Y2 - 22 July 2022 through 24 July 2022

ER -

Su W, Li Q. An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines. In 2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2022. (2022 5th International Conference on Data Science and Information Technology, DSIT 2022 - Proceedings). doi: 10.1109/DSIT55514.2022.9943927

An Efficient Method for Sample Adversarial Perturbations against Nonlinear Support Vector Machines

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this