Employing SAE-GRU deep learning for scalable botnet detection in smart city infrastructure

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Smart City Technology

Background:

• Internet of Things (IoT) devices in smart cities increase urban infrastructure capabilities but also elevate botnet attack risks.
• Botnet attacks pose significant threats to essential services and public safety in smart city environments.
• Existing intrusion detection systems struggle with the high-dimensional data and temporal patterns characteristic of IoT networks.

Purpose of the Study:

• To develop a novel hybrid deep learning model for real-time botnet detection and mitigation in smart city IoT networks.
• To address the challenges of processing high-dimensional data and recognizing temporal patterns in IoT security.
• To offer a scalable and resource-efficient solution for enhancing the security of smart city ecosystems.

Main Methods:

• A hybrid deep learning model, Stacked Autoencoder-Gated Recurrent Unit (SAE-GRU), was developed.
• Stacked Autoencoders were used for data dimensionality reduction.
• Gated Recurrent Units were employed for analyzing sequential data patterns.
• An emulated smart city environment served as a testbed for model evaluation.

Main Results:

• The SAE-GRU model demonstrated significant improvements in botnet detection performance.
• Average accuracy reached 98.65 percent, with consistently high precision and recall values.
• The model proved effective in identifying and mitigating botnet activities in a realistic smart city setting.

Conclusions:

• The SAE-GRU model offers a scalable and resource-efficient solution for botnet detection in smart city IoT networks.
• The findings contribute to a better understanding of IoT security challenges and deep learning applications.
• This research lays the groundwork for adaptive security mechanisms against emerging threats in smart cities.