Research on Network Attack Sample Generation and Defence Techniques Based on Generative Adversarial Networks DOI Open Access
Jing Shan, Hong Ma, Jian Li

et al.

Applied Mathematics and Nonlinear Sciences, Journal Year: 2024, Volume and Issue: 9(1)

Published: Jan. 1, 2024

Abstract Generative Adversarial Networks, as a powerful generative model, show great potential in generating adversarial samples and defending against attacks. In this paper, using Networks (GANs) the basic framework, we design network attack sample generation method based on Deep Convolutional (DCGANs) an defence multi-scale GANs, verify practicality of two methods through experiments, respectively. Compared with three AE-CDA, AE-DEEP AE-ATTACK, DCGAN-based paper can interfere detection function anomaly model more effectively, has better stability versatility, maintain relatively stable effect wide range models datasets. On MNIST dataset, classification accuracy proposed is only slightly lower than that APE-GAN JSMA samples, maximum 98.69%. The reaches 98.69%, time consumption 1.5 s, which larger 1.2 s. Thus, paper’s GAN-based defense smaller or equal to other comparative when systematic errors are ignored. purpose provide technical reference how eliminate perturbations networks.

Language: Английский

Ancient Book Image Restoration Using Generative Adversarial Networks DOI Creative Commons

Nan Wang,

Jiajian Zhu,

Shang Shi

et al.

Highlights in Science Engineering and Technology, Journal Year: 2025, Volume and Issue: 133, P. 128 - 136

Published: Feb. 25, 2025

As an important carrier of historical and cultural inheritance, the restoration ancient books is great significance to protection relics inheritance. However, traditional repair methods have some problems, such as low efficiency insufficient precision. In this paper, a deep learning-based method for proposed, which divided into two steps: structure reconstruction color correction. The network (SRN) uses line drawing information ensure authenticity structural stability large-scale content, correction (CCN) makes local adjustments missing pixels, reducing bias edge hopping problems. experimental results show that effectively improves image quality, provides new technical support inheritance books.

Language: Английский

Citations

0
Optimized Bayesian tensorized neural network affording task failure prediction in cloud environment DOI

Senthil Kumar Avinashi Malleswaran,

Kalimuthu Marimuthu,

Philippe Robert

et al.

Expert Systems with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 127538 - 127538

Published: April 1, 2025

Language: Английский

Citations

0
Research on Network Attack Sample Generation and Defence Techniques Based on Generative Adversarial Networks DOI Open Access
Jing Shan, Hong Ma, Jian Li

et al.

Applied Mathematics and Nonlinear Sciences, Journal Year: 2024, Volume and Issue: 9(1)

Published: Jan. 1, 2024

Abstract Generative Adversarial Networks, as a powerful generative model, show great potential in generating adversarial samples and defending against attacks. In this paper, using Networks (GANs) the basic framework, we design network attack sample generation method based on Deep Convolutional (DCGANs) an defence multi-scale GANs, verify practicality of two methods through experiments, respectively. Compared with three AE-CDA, AE-DEEP AE-ATTACK, DCGAN-based paper can interfere detection function anomaly model more effectively, has better stability versatility, maintain relatively stable effect wide range models datasets. On MNIST dataset, classification accuracy proposed is only slightly lower than that APE-GAN JSMA samples, maximum 98.69%. The reaches 98.69%, time consumption 1.5 s, which larger 1.2 s. Thus, paper’s GAN-based defense smaller or equal to other comparative when systematic errors are ignored. purpose provide technical reference how eliminate perturbations networks.

Language: Английский

Citations

0