<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>We urgently need to explore the communication security guarantee mechanism to ensure the security of transmitting important information, and the steganographic communication model plays a crucial role in achieving this security. In this paper, we focus our research on secure communication methods using images, a widely used multimedia transmission medium. We design two covert communication models, LSTM-ACGAN and CNN-GAN, to enhance the security of deep image steganography algorithms while also improving their generalization ability. This paper’s model is compared with three current mainstream steganographic communication models on a dataset, and the CNN-GAN model is applied to a confidential unit dataset transmission. The research reveals that the two models in this paper possess the ability to detect across multiple domains, remain unaffected by specific domains, and require only 30%–33% of the parameters and arithmetic operations compared to mainstream models. Furthermore, they exhibit greater versatility and practicability, with training accuracy reaching 90% or higher. The model presented in this paper can meet the computer information security communication task after several rounds of iteration and meets the design expectations. This study provides a useful exploration for the establishment of computer information security guarantee mechanisms and designs a proven secure covert communication scheme.</p>
</abstract>

We urgently need to explore the communication security guarantee mechanism to ensure the security of transmitting important information, and the steganographic communication model plays a crucial role in achieving this security. In this paper, we focus our research on secure communication methods using images, a widely used multimedia transmission medium. We design two covert communication models, LSTM-ACGAN and CNN-GAN, to enhance the security of deep image steganography algorithms while also improving their generalization ability. This paper’s model is compared with three current mainstream steganographic communication models on a dataset, and the CNN-GAN model is applied to a confidential unit dataset transmission. The research reveals that the two models in this paper possess the ability to detect across multiple domains, remain unaffected by specific domains, and require only 30%–33% of the parameters and arithmetic operations compared to mainstream models. Furthermore, they exhibit greater versatility and practicability, with training accuracy reaching 90% or higher. The model presented in this paper can meet the computer information security communication task after several rounds of iteration and meets the design expectations. This study provides a useful exploration for the establishment of computer information security guarantee mechanisms and designs a proven secure covert communication scheme.

A Computerized Information Security Communication Method Based on Deep Learning Theory

Fair Friend Institute of Intelligent Manufacturing

Applied Mathematics and Nonlinear Sciences

This work is licensed under the Creative Commons Attribution 4.0 International License.

We urgently need to explore the communication security guarantee mechanism to ensure the security of transmitting important information, and the steganographic...

A Computerized Information Security Communication Method Based on Deep Learning Theory

Online veröffentlicht: 05. Aug. 2024

Seitenbereich: -

Eingereicht: 05. Apr. 2024

Akzeptiert: 04. Juli 2024

DOI: https://doi.org/10.2478/amns-2024-2249

Schlüsselwörter
LSTM, GAN, CNN, High frequency residual features, Covert communication

© 2024 Yingyun Kang, published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

A Computerized Information Security Communication Method Based on Deep Learning Theory

Online veröffentlicht: 05. Aug. 2024

Seitenbereich: -

Eingereicht: 05. Apr. 2024

Akzeptiert: 04. Juli 2024

DOI: https://doi.org/10.2478/amns-2024-2249

SchlüsselwörterLSTM, GAN, CNN, High frequency residual features, Covert communication

© 2024 Yingyun Kang, published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Schlüsselwörter
LSTM, GAN, CNN, High frequency residual features, Covert communication