Lightweight TinyML NIDS for IoT devices

With the increase in the implementation of Internet of Things (IoT) devices, this has introduced some serious and critical security challenges, specifically with regards to the resource-constrained environments, which are IoT devices that are used by the public for convenience. Furthermore, traditional intrusion detection systems (IDS) are too computationally intensive for such devices, emphasizing on the need for a much lightweight and efficient alternatives. This paper focuses on developing a lightweight IDS using TinyML to tackle these security challenges and ensure safety in IoT devices. Public IoT datasets were preprocessed to enhance the quality of datasets, and it is followed by feature importance and class distribution analyses. The machine learning models, K-Nearest Neighbors (KNN), Decision Tree, and Logistic Regression, were trained for intrusion detection. The models were then converted to C++ code format for deployment on a Raspberry Pi simulation. Testing and evaluation were conducted using opensource Quick Emulator (QEMU), which uses dynamic binary translation to emulate a computer’s processor. QEMU provides a variety of hardware and device models for virtual machines, enabling it to run different guest operating systems, making it an ideal option. The testing and evaluation processes were conducted to assess performance using metrics like accuracy, precision, recall, and resource utilization. The results demonstrate that the proposed system achieves high detection accuracy with minimal computational overhead, making it a viable solution for securing resource-constrained IoT devices. This study underscores the potential of TinyML in improving and enhancing IoT security while ensuring scalability as well as efficiency.