کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Real-Time FPGA Implementation of a Hybrid Random Forest–MLP Model for Intrusion Detection

نویسندگان :

Meysam Kiani¹ Hamid Shafaghi² Siavash Es’haghi³

1- dept.of electrical engineering SR.C., Islamicname Azad University Tehran, Iran 2- Faculty of Electrical Engineering, Shahid Beheshti University 3- dept.of electrical engineering SR.C., Islamicname Azad University Tehran, Iran

کلمات کلیدی :

Intrusion Detection System،Multi-Layer Perceptron،FPGA،Low-Power،Real-Time،Network Security

چکیده :

Today, communication and Internet networks have expanded significantly. Given their increasing applications, ensuring the security of these networks has become essential. Consequently, numerous studies have been conducted on Intrusion Detection Systems (IDSs). In this study, a hybrid IDS combining the Random Forest algorithm and a multilayer perceptron (MLP) neural network is proposed. The Random Forest method is employed to select the most relevant and effective features as inputs to the neural network, reducing input dimensionality, increasing learning speed, and enhancing model performance. The proposed model is evaluated using the NSL-KDD dataset with k-fold cross-validation, achieving an accuracy of 98.92%. Furthermore, the MLP model is implemented on an FPGA by applying fixed weights and utilizing efficient multipliers. The hardware system operates with a latency of 23 clock cycles and functions in real time. Hardware optimizations eliminate the use of DSPs and BRAMs, reducing resource consumption and system power. The dynamic power consumption of the system is measured at 886 mW. Based on the obtained results, the proposed IDS represents a highly efficient, low-power, and real-time solution for network security applications, particularly suitable for deployment in high-speed networks and critical infrastructure systems.