A two-tier optimization strategy for feature selection in robust adversarial attack mitigation on internet of things...

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Internet of Things (IoT)

Background:

• Adversarial attacks pose a growing threat to network security, particularly in the context of interconnected IoT systems.
• While Deep Learning (DL) models are used in Intrusion Detection Systems (IDS), their vulnerability to adversarial examples remains an open research area.
• The increasing reliance on IoT, AI, and 5G in Industry 4.0 amplifies security concerns due to vast data processing.

Purpose of the Study:

• To introduce a novel Two-Tier Optimization Strategy for Robust Adversarial Attack Mitigation (TTOS-RAAM) model for enhanced IoT network security.
• To address the challenge of detecting adversarial attack behavior within IoT environments.
• To evaluate the effectiveness of DL models against adversarial attacks in IoT networks.

Main Methods:

• Data preprocessing using a min-max scaler for uniform input.
• Optimal feature selection employing a hybrid of Coati-Grey Wolf Optimization (CGWO).
• Adversarial attack detection using a Conditional Variational Autoencoder (CVAE), with parameter tuning via Improved Chaos African Vulture Optimization (ICAVO).

Main Results:

• The TTOS-RAAM model demonstrated superior performance in detecting adversarial attacks.
• Experimental analysis on the RT-IoT2022 dataset showed a high accuracy of 99.91%.
• The proposed method significantly outperforms existing approaches in adversarial attack mitigation for IoT networks.

Conclusions:

• The TTOS-RAAM model provides a robust solution for mitigating adversarial attacks in IoT network security.
• The study highlights the potential of advanced optimization and DL techniques for securing IoT data.
• The findings contribute to the ongoing research on defending intelligent systems against sophisticated cyber threats.