Effective Feature Selection Methods to Detect IoT DDoS Attack in 5G Core Network.

Area of Science:

• Network Security
• Machine Learning
• 5G Communications

Background:

• 5G networks enable massive Internet of Things (IoT) but face security risks from vulnerable IoT devices.
• Weak IoT security can lead to terabits per second (Tbps) Distributed Denial of Service (DDoS) attacks on 5G infrastructure.
• Existing machine learning (ML) DDoS detection models are primarily designed for wired networks, with limited research on 5G traffic.

Purpose of the Study:

• To investigate the effectiveness of feature selection in reducing the time complexity of ML-based DDoS attack detection in 5G core networks.
• To improve the real-time detection capabilities for large-capacity DDoS attacks within 5G environments.
• To address the insufficiency of feature engineering studies specific to 5G network traffic.

Main Methods:

• Conducted feature selection experiments on large datasets within a 5G core network environment.
• Applied ML techniques to detect and analyze DDoS attacks in real-time.
• Evaluated the impact of feature selection on detection performance and time complexity.

Main Results:

• Feature selection maintained or improved the performance of ML-based DDoS detection models.
• The time complexity reduction achieved through feature selection increased significantly with larger datasets.
• Real-time detection of large-scale DDoS attacks in 5G core networks was demonstrated as feasible using feature selection.

Conclusions:

• Feature selection is crucial for removing noisy features, thereby enhancing the efficiency and performance of ML models for DDoS attack detection.
• The study confirms the viability of ML with feature selection for low-latency, real-time DDoS attack detection in 5G networks.
• This research contributes to advancing automated AI-driven DDoS attack detection technologies for 5G networks.