A multilayer deep autoencoder approach for cross layer IoT attack detection using deep learning algorithms.

Area of Science:

• Cybersecurity
• Network Security
• Internet of Things (IoT)

Background:

• Cybersecurity professionals require advanced techniques to detect subtle anomalies in complex network data.
• Modern threat landscapes necessitate improved methods for feature representation, scalability, and flexibility in cybersecurity solutions.

Purpose of the Study:

• To introduce the Multi-Layer Deep Autoencoder (M-LDAE) for cross-layer Internet of Things (IoT) threat detection.
• To address challenges in feature representation, scalability, and flexibility in current cybersecurity techniques.

Main Methods:

• Utilized deep autoencoders' hierarchical simplification for extracting global and local attributes.
• Integrated deep learning algorithms like Recurrent Neural Networks (RNNs), Graph Neural Networks (GNNs), and Temporal Convolutional Networks (TCNs).
• Employed benchmark datasets and real-world scenarios for extensive simulations.

Main Results:

• M-LDAE effectively safeguards against Man-in-the-Middle (MitM) and Distributed Denial of Service (DDoS) attacks in IoT networks.
• Demonstrated adaptability to new attack vectors and improved detection accuracy.
• Significantly reduced false positive rates in threat detection.

Conclusions:

• M-LDAE presents a novel paradigm for cross-layer IoT attack detection, offering a flexible and robust cybersecurity solution.
• The proposed method enhances cyber threat identification across diverse IoT domains.
• M-LDAE improves overall cybersecurity resilience in the evolving threat environment.