Isolation Forest Based TinyML for Detecting Hardware Trojans on FPGA in Real Time

The growing utilization of microcontrollers and FPGAs across several sectors has amplified the vulnerability to hardware attacks, namely the potential for information leakage via hardware Trojans. Real-time detection of these Trojans is essential for preserving the integrity and security of hardware systems. This research presents a new method for detecting hardware Trojans in real-time by combining TinyML and FPGA technology. By utilizing the capabilities of FPGA, we gather timing and power data to detect irregularities that could potentially indicate the existence of hardware Trojans. We employ a training process that utilizes an Isolation Forest machine learning model, which is recognized for its strong capability in identifying anomalies. Subsequently, the model is transformed into Verilog code and implemented on the FPGA, facilitating the detection of Trojans with high efficiency and effectiveness during the execution of a program. The proposed solution not only improves the security of the hardware, but also proves the possibility of combining advanced machine learning models with FPGA technology for real-time applications. The outcomes of our study demonstrate substantial enhancements in the precision and efficiency of detection, therefore facilitating the development and implementation of more secure hardware designs.