Digital security risk identification and model construction of smart city based on deep learning

Area of Science:

• Computer Science
• Cybersecurity
• Smart City Technology

Background:

• Smart cities increasingly integrate the Industrial Internet of Things (IIoT), expanding the attack surface and introducing significant network security risks.
• Existing security measures struggle to cope with the complexity and scale of IIoT in smart city environments.

Purpose of the Study:

• To propose a novel deep learning-based digital security identification model (DL-DSIM) for enhancing IIoT security in smart cities.
• To improve data transmission efficiency and overall system security within smart city infrastructures.

Main Methods:

• A flexible three-layer architecture framework was designed for the DL-DSIM.
• A hybrid feature selection method combining CSO and GA was introduced to reduce complexity.
• Deep Neural Networks (DNNs) were employed for enhanced intrusion detection and vulnerability processing.

Main Results:

• DL-DSIM achieved high performance metrics in the training phase: 99.13% accuracy, 98.5% recall, 98.39% F-value, and 95.62% specificity.
• In the test phase, DL-DSIM demonstrated strong performance with 96.1% accuracy, 95.48% recall, 96.38% F-value, and 93% specificity.
• The model proved efficient in resisting network security threats.

Conclusions:

• DL-DSIM offers a reliable security mechanism for IIoT systems in smart cities.
• The proposed model contributes to the sustainable development of smart cities and digital security infrastructure.
• The research highlights the effectiveness of deep learning in addressing complex cybersecurity challenges in urban environments.