AI-Based Malicious Encrypted Traffic Detection in 5G Data Collection and Secure Sharing DOI Open Access
Gang Han, Hailin Zhang, Zhong-Liang Zhang

et al.

Electronics, Journal Year: 2024, Volume and Issue: 14(1), P. 51 - 51

Published: Dec. 26, 2024

With the development and widespread application of network information, new technologies led by 5G are emerging, resulting in an increasingly complex security environment more diverse attack methods. Unlike traditional networks, networks feature higher connection density, faster data transmission speeds, lower latency, which widely applied scenarios such as smart cities, Internet Things, autonomous driving. The vast amounts sensitive generated these applications become primary targets during processes collection secure sharing, unauthorized access or tampering could lead to severe breaches integrity issues. However, extensively employ encryption protect transmission, attackers can hide malicious content within encrypted communication, rendering content-based traffic detection methods ineffective for identifying traffic. To address this challenge, paper proposes a method based on reconstructive domain adaptation adversarial hybrid neural networks. proposed integrates generative with ResNet, ResNeXt, DenseNet construct network, aiming tackle challenges detection. On basis, module is introduced reduce distribution discrepancy between source target domain, thereby enhancing cross-domain capabilities. By preprocessing from public datasets, capable extracting deep features without need decryption. generator utilizes generate realistic samples, while discriminator achieves sample classification through high-dimensional extraction. Additionally, classifier further improves model’s stability generalization across different environments time periods. Experimental results demonstrate that significantly accuracy efficiency environments, effectively performance

Language: Английский

A Study on the Construction of Mental Health Indicators for College Students Based on Social Media Data Mining and the Evaluation of Their Intervention Effects DOI Open Access

Huaichen Ji

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

Published: Jan. 1, 2024

Abstract The development of social media has brought many tests to the mental health education college students, and some students have fallen into network addiction dependence, which greatly affects their physical health. article uses microblogging as source students’ data preprocesses using de-emphasis Chinese word separation. It also analyzes problematic manifestations in colleges universities, extracts indicators by TF-IDF algorithm, realizes recognition topics BTM model. CNN-LSTM-ATT model was established introducing attention mechanism LSTM assess status students. analyzed terms characteristics predictive validation used develop intervention strategies for text length is [1,22], occupies 86.98% all sentences, AUC value corresponding 0.946, prediction accuracy CNN-LSTMATT assessment universities can reach up 97.62%. clarify realize construction from dimensions literacy regulatory mechanisms.

Language: Английский

Citations

1

A multiscale approach for network intrusion detection based on variance–covariance subspace distance and EQL v2 DOI
Can Liu, Yu Fu, Kun Wang

et al.

Computers & Security, Journal Year: 2024, Volume and Issue: unknown, P. 104173 - 104173

Published: Oct. 1, 2024

Language: Английский

Citations

0

AI-Based Malicious Encrypted Traffic Detection in 5G Data Collection and Secure Sharing DOI Open Access
Gang Han, Hailin Zhang, Zhong-Liang Zhang

et al.

Electronics, Journal Year: 2024, Volume and Issue: 14(1), P. 51 - 51

Published: Dec. 26, 2024

With the development and widespread application of network information, new technologies led by 5G are emerging, resulting in an increasingly complex security environment more diverse attack methods. Unlike traditional networks, networks feature higher connection density, faster data transmission speeds, lower latency, which widely applied scenarios such as smart cities, Internet Things, autonomous driving. The vast amounts sensitive generated these applications become primary targets during processes collection secure sharing, unauthorized access or tampering could lead to severe breaches integrity issues. However, extensively employ encryption protect transmission, attackers can hide malicious content within encrypted communication, rendering content-based traffic detection methods ineffective for identifying traffic. To address this challenge, paper proposes a method based on reconstructive domain adaptation adversarial hybrid neural networks. proposed integrates generative with ResNet, ResNeXt, DenseNet construct network, aiming tackle challenges detection. On basis, module is introduced reduce distribution discrepancy between source target domain, thereby enhancing cross-domain capabilities. By preprocessing from public datasets, capable extracting deep features without need decryption. generator utilizes generate realistic samples, while discriminator achieves sample classification through high-dimensional extraction. Additionally, classifier further improves model’s stability generalization across different environments time periods. Experimental results demonstrate that significantly accuracy efficiency environments, effectively performance

Language: Английский

Citations

0