FPGA/AI-Powered Architecture for Anomaly Network Intrusion Detection Systems

This paper proposes an architecture to develop machine learning/deep learning models for anomaly network intrusion detection systems on reconfigurable computing platforms. We build two validate the framework: Anomaly Detection Autoencoder (ADA) and Artificial Neural Classification (ANC) in NetFPGA-sume platform. Three published data sets NSL-KDD, UNSW-NB15, CIC-IDS2017 are used test deployed models’ throughput, latency, accuracy. Experimental results with NetFPGA-SUME show that ADA model uses 20.97% LUTs, 15.16% FFs, 19.42% BRAM, 6.81% DSP while ANC requires 21.39% 15.19% FFS, 14.59% 3.67% DSP. achieve a bandwidth of up 28.7 Gbps 34.74 Gbps, respectively. In terms can process at 18.7 Gops, offer 10 Gops different datasets. With NSL-KDD dataset, achieves 90.87% accuracy false negative rate 4.86%. The UNSW-NB15 obtains 87.49% 98.22%, respectively, rates achieving 2.0% 6.2%,