AI-Enabled Digital Twin Framework for TSCA-like Anomaly Detection in FPGA-SoC-Based Industrial Cyber-Physical Systems

Area of Science:

• Cyber-Physical Systems (CPSs)
• Hardware Security
• Integrated Circuit Design

Background:

• Field-Programmable Gate Array System-on-Chip (FPGA-SoC) platforms are vital in industrial CPSs.
• Increased integration density in FPGA-SoCs creates thermal security vulnerabilities.
• Thermal Side-Channel Attacks (TSCAs) exploit heat variations for information leakage.

Purpose of the Study:

• To develop an AI-enabled Digital Twin (DT) framework for detecting TSCAs in FPGA-SoC microarchitectures.
• To advance beyond conventional Thermal Digital Twin (TDT) approaches for autonomous threat detection.
• To address critical hardware security challenges in next-generation industrial CPS infrastructures.

Main Methods:

• Developed an AI-enabled Digital Twin (DT) framework.
• Combined thermal behavioral modeling, feature engineering, and machine learning-based anomaly detection.
• Validated the framework on an NI myRIO-1900 platform with a Xilinx Zynq-7010 FPGA-SoC.

Main Results:

• Achieved approximately 75% accuracy in TSCA detection.
• Obtained an Area Under the ROC Curve (AUC) of 0.76 using an Isolation Forest model.
• Demonstrated the framework's feasibility in learning normal thermal patterns and detecting anomalies.

Conclusions:

• The AI-enabled DT framework effectively detects anomalous thermal activities indicative of TSCAs.
• This approach enhances the security of densely integrated FPGA-SoC platforms in industrial CPSs.
• The study validates the capability of AI-driven DTs for autonomous hardware security threat detection.