Cross-Layer Analysis of Machine Learning Models for Secure and Energy-Efficient IoT Networks

Area of Science:

• Computer Science
• Electrical Engineering
• Cybersecurity

Background:

• Widespread Internet of Things (IoT) adoption presents significant security and energy efficiency challenges, especially for resource-constrained devices.
• Existing IoT security solutions often struggle to balance robust protection with minimal power consumption.

Purpose of the Study:

• To propose a novel cross-layer IoT architecture that integrates machine learning (ML) models and lightweight cryptography.
• To enhance security, improve authentication, and optimize energy efficiency in large-scale IoT deployments.

Main Methods:

• Implementation of a cross-layer IoT architecture featuring role-based access control (RBAC) with energy-aware policies.
• Integration of layer-specific ML models (LSTM for anomaly detection, decision trees for validation) and adaptive Speck encryption.
• Leveraging convolutional neural networks (CNNs) for enhanced IoT security and energy efficiency.

Main Results:

• Reduced false positives by up to 32%.
• Prevented unauthorized access attempts with up to 95% effectiveness.
• Achieved a 30% reduction in power consumption using lightweight Speck encryption compared to AES.

Conclusions:

• The proposed cross-layer IoT architecture effectively harmonizes ML-driven security with energy-efficient operations.
• The system demonstrates significant improvements in security and power consumption for practical IoT applications.
• This approach offers a viable solution for securing smart cities, homes, and schools.