Kun Liu
, En Zeng, Bohan Liu, Junda Li, Jiangrong Li
To ensure the security of the attack detection model of time series data, an adversarial attack and adversarial defense method based on multivariate time series data was proposed. First, the escape attack implemented in the test phase was designed for the autoencoder-based attack detection model. Second, according to the designed adversarial attack samples, the adversarial defense strategy based on the Jacobian regularization method was proposed. The Jacobian matrix in the calculation model training process was taken as the regular term in the objective function to improve the defense capability of the deep learning model. The attack effects of the proposed attack methods and the defense effect of the proposed adversarial defense method were verified on the BATADAL dataset of industrial water treatment.
Liu, K
, Zeng, E, Liu, B, Li, J & Li, J 2023, '
基于多变量时序数据的对抗攻击与防御方法
',
Beijing Gongye Daxue Xuebao / Journal of Beijing University of Technology
, 卷 49, 号码 4, 页码 415-423.
https://doi.org/10.11936/bjutxb2022090028
© 2023 Beijing University of Technology. All rights reserved.
PY - 2023/4
Y1 - 2023/4
N2 - To ensure the security of the attack detection model of time series data, an adversarial attack and adversarial defense method based on multivariate time series data was proposed. First, the escape attack implemented in the test phase was designed for the autoencoder-based attack detection model. Second, according to the designed adversarial attack samples, the adversarial defense strategy based on the Jacobian regularization method was proposed. The Jacobian matrix in the calculation model training process was taken as the regular term in the objective function to improve the defense capability of the deep learning model. The attack effects of the proposed attack methods and the defense effect of the proposed adversarial defense method were verified on the BATADAL dataset of industrial water treatment.
AB - To ensure the security of the attack detection model of time series data, an adversarial attack and adversarial defense method based on multivariate time series data was proposed. First, the escape attack implemented in the test phase was designed for the autoencoder-based attack detection model. Second, according to the designed adversarial attack samples, the adversarial defense strategy based on the Jacobian regularization method was proposed. The Jacobian matrix in the calculation model training process was taken as the regular term in the objective function to improve the defense capability of the deep learning model. The attack effects of the proposed attack methods and the defense effect of the proposed adversarial defense method were verified on the BATADAL dataset of industrial water treatment.
KW - Jacobian regularization
KW - adversarial attack
KW - adversarial defense
KW - attack detection
KW - autoencoder
KW - multivariate time series
UR - http://www.scopus.com/inward/record.url?scp=85153099162&partnerID=8YFLogxK
U2 - 10.11936/bjutxb2022090028
DO - 10.11936/bjutxb2022090028
M3 - 文章
AN - SCOPUS:85153099162
SN - 0254-0037
VL - 49
SP - 415
EP - 423
JO - Beijing Gongye Daxue Xuebao / Journal of Beijing University of Technology
JF - Beijing Gongye Daxue Xuebao / Journal of Beijing University of Technology
IS - 4
ER -
我们使用 cookie 来提供和增强我们的服务以及定制内容。如果继续，则表示您同意
cookie 的使用
