Combination of quantum-based optimizer and feature pyramid network for intrusion detection in Cloud-IoT environments

Area of Science:

• Cybersecurity
• Network Intrusion Detection Systems (IDS)
• Cloud-IoT Security

Background:

• The dynamic Cloud-IoT environment necessitates advanced intrusion detection systems (IDS).
• Existing solutions struggle with high-dimensional, heterogeneous network traffic management.
• Limitations in current IDS hinder effective threat detection in complex network infrastructures.

Purpose of the Study:

• To present an improved IDS structure for Cloud-IoT environments.
• To enhance the management of high-dimensional and heterogeneous network traffic.
• To leverage quantum-inspired optimization for superior IDS performance.

Main Methods:

• Developed an Enhanced Feature Pyramid Network (EFPN) adapted for tabular network data using multi-scale feature extraction.
• Integrated a Quantum-Enhanced Child Drawing Development Optimizer (Q-CDDO) utilizing quantum rotation gates for hyperparameter tuning.
• Validated the model on benchmark datasets: CIC-IDS-2017 and Bot-IoT.

Main Results:

• Achieved 96.3% accuracy on the CIC-IDS-2017 dataset.
• Achieved 94.6% accuracy on the Bot-IoT dataset.
• Ablation studies confirmed the synergistic effect of EFPN and Q-CDDO; visualization verified the model's discriminative power.

Conclusions:

• The proposed EFPN and Q-CDDO model significantly advances intrusion detection capabilities in Cloud-IoT.
• Quantum-inspired metaheuristics demonstrate strong potential for enhancing cybersecurity solutions.
• The study validates the effectiveness of the integrated approach for dynamic network traffic analysis.