Artificial intelligence-driven cybersecurity system for internet of things using self-attention deep learning and...

Area of Science:

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

Background:

• Rapid expansion of Internet of Things (IoT) devices has led to increased cybersecurity vulnerabilities.
• Cyberattacks on IoT are unique due to device limitations, necessitating specialized defense mechanisms.
• Artificial Intelligence (AI) is increasingly used in cybersecurity, but adversarial AI poses new threats.

Purpose of the Study:

• To propose an Intelligent Cybersecurity System Using Self-Attention-based Deep Learning and Metaheuristic Optimization Algorithm (ICSSADL-MHOA) for robust IoT security.
• To enhance the detection and classification of cybersecurity threats in IoT environments.
• To develop an adaptive and real-time cybersecurity model against advanced AI-driven attacks.

Main Methods:

• Data normalization using min-max normalization for consistency.
• Feature selection via improved tuna swarm optimization (ITSO).
• Cybersecurity threat detection and classification using bidirectional long short-term memory with self-attention (BiLSTM-SA), optimized by hunger games search (HGS).

Main Results:

• The ICSSADL-MHOA model demonstrated superior performance on ToN-IoT and Edge-IIoT datasets.
• Achieved a high accuracy of 99.37% in detecting and classifying cybersecurity threats.
• Outperformed existing cybersecurity techniques in experimental validation.

Conclusions:

• The proposed ICSSADL-MHOA model offers an effective solution for enhancing IoT cybersecurity.
• The integration of deep learning and metaheuristic algorithms provides a robust defense against sophisticated cyber threats.
• This research contributes to developing advanced, adaptive cybersecurity systems for the evolving threat landscape.