NGMD: next generation malware detection in federated server with deep neural network model for autonomous networks

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Network Security

Background:

• Distributed denial-of-service (DDoS) attacks pose a significant threat to internet users and providers.
• Traditional signature-based detection systems struggle against evolving malware.
• Deep learning (DL) offers advanced solutions for dynamic threat detection, especially in cloud-edge and federated environments.

Purpose of the Study:

• To deploy an optimized deep neural network for network traffic classification.
• To coordinate federated server model parameters with IoT device training.
• To develop a privacy-preserving federated learning framework for enhanced DDoS attack detection.

Main Methods:

• Implemented a federated learning approach for privacy-aware model training across distributed IoT devices.
• Utilized an optimized deep neural network for network traffic classification.
• Developed a federated flowchart for local model updates aggregation and global model generation at the cloud-edge.

Main Results:

• The federated learning model demonstrated reliable attack detection with efficient classification, privacy, and confidentiality.
• The proposed framework requires minimal memory and network delay.
• Experimental validation on the BoT-IoT dataset confirmed superior performance compared to centralized and localized DL models.

Conclusions:

• Federated learning provides a robust and privacy-preserving solution for DDoS attack detection in IoT.
• The developed framework enhances network resilience and security in autonomous networks.
• This approach facilitates collaborative, privacy-aware learning without compromising user data.