A GAN-Based Approach for enhancing security in satellite based IoT networks using MPI enabled HPC

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Satellite Communications

Background:

• Satellite Internet of Things (IoT) networks are crucial for critical applications but face security vulnerabilities due to diverse technologies and limited device capacity.
• Secure data transmission is a primary concern when connecting IoT systems to High-Performance Computing (HPC) clouds via satellite links.

Purpose of the Study:

• To propose a novel security framework, DLGAN (Deep Learning-based Generative Adversarial Network), specifically for satellite-based IoT environments.
• To address the challenge of skewed datasets in cybersecurity by generating synthetic attack data using Generative Adversarial Networks (GANs).
• To enable scalable parallel processing of large IoT data volumes on HPC systems using the Message Passing Interface (MPI).

Main Methods:

• Utilized Convolutional Neural Networks (CNNs) for real-time anomaly detection.
• Employed Generative Adversarial Networks (GANs) to create realistic synthetic attack data.
• Implemented a generator/discriminator mechanism for classifying network traffic as benign or malicious.
• Optimized the DLGAN model for HPC systems with AI-enabled GPUs for efficient parallel processing.

Main Results:

• The DLGAN framework demonstrated enhanced detection accuracy for 14 different attack types.
• Achieved significant reductions in model training time.
• Showcased excellent scalability with large data volumes, suitable for real-time security operations.
• Maintained low computational costs while providing fast and accurate threat detection.

Conclusions:

• Integrating deep learning with HPC-based distributed environments offers an efficient and dynamic defense for IoT networks.
• The DLGAN solution provides a scalable, efficient, and attack-resilient mechanism for securing satellite-based IoT infrastructures.
• The framework effectively addresses the unique security challenges of satellite IoT networks.