Anomaly-Based Intrusion Detection System Using ESP32-WROOM-DA

Internet of Things (IoT) devices are increasingly employed in monitoring and controlling both domestic and industrial infrastructures. However, security measures are often neglected due to the computational resource limitations of these devices. Despite numerous research initiatives aimed at developing Intrusion Detection Systems (IDS) for IoT, practical implementation-focused studies remain scarce. The goal of this research is to develop an anomaly-based IDS using a supervised approach with three different neural network models: Sequential Neural Network (SNN), Recurrent Neural Network (RNN), and Deep Recurrent Neural Network (DRNN). The objective is to determine whether it is feasible to create and deploy a high-performing IDS directly on the ESP32 board while simultaneously maintaining low resource requirements. To achieve this, the IDS is first trained and then tested on the NSL-KDD dataset. Results show that the most accurate IDS utilizes the SNN model, achieving a precision level of 94.04%. This IDS, when deployed on the ESP32-WROOM-32 microcontroller, reports a minimum inference time of 0.226 ms, an average time of 3.198 ms, and a maximum time of 10.478 ms, requiring just over 8 KB of SRAM for installation.