Augmenting cybersecurity through attention based stacked autoencoder with optimization algorithm for detection and...

Area of Science:

• Cybersecurity
• Network Security
• Machine Learning in IoT

Background:

• Internet of Things (IoT) networks face increasing cybersecurity threats due to massive connectivity.
• Traditional Intrusion Detection Systems (IDS) struggle with the scale and diversity of IoT devices.
• Machine Learning (ML) and Deep Learning (DL) offer promising solutions for IoT security challenges.

Purpose of the Study:

• To propose a novel methodology, CASAE-POADMA, for detecting and mitigating cybersecurity attacks in IoT networks.
• To enhance the security posture of IoT-assisted environments through advanced computational techniques.
• To address the limitations of existing security measures in the context of evolving IoT threats.

Main Methods:

• The CASAE-POADMA methodology employs min-max normalization for data scaling.
• Feature selection is performed using the greylag goose optimization (GGO) method.
• Cybersecurity attack detection and mitigation are achieved using an attention-based stacked autoencoder (ASAE), with hyperparameter tuning via the pelican optimization algorithm (POA).

Main Results:

• The CASAE-POADMA approach demonstrated superior performance in identifying and mitigating cybersecurity attacks.
• Experimental validation on a benchmark database yielded a high accuracy of 99.50%.
• The proposed method significantly outperforms existing techniques in securing IoT networks.

Conclusions:

• The CASAE-POADMA methodology provides an effective solution for enhancing cybersecurity in IoT networks.
• The integration of attention-based stacked autoencoders and pelican optimization algorithm offers a robust framework for attack detection and mitigation.
• The high accuracy achieved validates the potential of this approach for real-world IoT security applications.