SecEdge: A novel deep learning framework for real-time cybersecurity in mobile IoT environments.

Area of Science:

• Cybersecurity
• Deep Learning
• Internet of Things (IoT)

Background:

• The proliferation of diverse, resource-constrained Internet of Things (IoT) devices poses significant cybersecurity challenges.
• Current security solutions struggle to address the dynamic and distributed nature of IoT environments.
• Real-time threat detection and data privacy are critical for mobile IoT systems.

Purpose of the Study:

• To propose SecEdge, a novel deep learning framework for enhanced real-time cybersecurity in mobile IoT environments.
• To integrate advanced AI techniques for robust threat detection and mitigation.
• To ensure data privacy and reduce latency in IoT security.

Main Methods:

• Integration of transformer-based models for dependency analysis and Graph Neural Networks (GNNs) for relational data.
• Implementation of federated learning for decentralized training and data privacy.
• Development of an adaptive learning mechanism for continuous model updates against evolving threats.

Main Results:

• SecEdge demonstrated superior performance across multiple benchmark datasets (NSL-KDD, UNSW-NB15, CICIDS2017).
• Achieved high detection rates: 98.8% for DoS attacks (NSL-KDD), 98.5% for MitM attacks (UNSW-NB15), and 98.7% for data injection attacks (CICIDS2017).
• Outperformed existing state-of-the-art cybersecurity methods in simulated environments.

Conclusions:

• SecEdge offers a promising solution for real-time cybersecurity in mobile IoT.
• The framework's hybrid deep learning approach effectively addresses complex cyber threats.
• Federated learning and adaptive mechanisms enhance the scalability and resilience of IoT security.