Machine Learning Methods to Detect Voltage Glitch Attacks on IoT/IIoT Infrastructures

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Modern IoT/IIoT systems depend on cryptography for security.
• Hardware attacks like voltage glitch attacks (VGAs) threaten system integrity and data confidentiality.
• VGAs can bypass security features or extract sensitive information by manipulating system voltage.

Purpose of the Study:

• To propose an innovative detection system for voltage glitch attacks (VGAs).
• To enhance the security of IoT/IIoT digital systems against hardware manipulation.
• To develop a robust defense mechanism using advanced machine learning techniques.

Main Methods:

• Implementation of a semisupervised learning methodology.
• Utilizing a hybrid combination of machine learning algorithms.
• Employing a heuristic clustering method with linear fragmentation for group classes.
• Leveraging Extreme Learning Machine (ELM) for hidden variable retrieval via convex optimization.

Main Results:

• Successful development of a novel VGA detection system.
• Demonstrated effectiveness of the semisupervised learning approach in identifying voltage manipulation.
• The hybrid algorithm combination proved adept at analyzing system behavior under attack.

Conclusions:

• The proposed machine learning system offers a promising solution for detecting VGAs in IoT/IIoT environments.
• Semisupervised learning, combined with heuristic clustering and ELM, provides a powerful framework for cybersecurity defense.
• This research contributes to securing digital infrastructure against sophisticated hardware-based attacks.