1 results
Search Results
Now showing 1 - 1 of 1
Master Thesis Radyo Frekansı (RF) Parmak İzi Kullanarak Cihaz Yetkilendirmesi(2024) İyiparlakoğlu, Raif; Dalveren, YaserWith the increasing usage areas of the Internet of Things (IoT), the importance of ensuring security in wireless networks has grown. Nevertheless, power-constrained devices cannot utilize intricate encryption techniques. Later on, Radio Frequency Fingerprinting (RFF) appeared, showcasing encouraging outcomes. A unique identity was derived from the distinct hardware variations during the production phases of IoT devices. This enabled device classification and verification functions, enhancing physical layer security. These applications were developed using deep learning (DL) techniques, resulting in highly accurate classification outcomes. Nevertheless, there is still room for enhancement when it comes to putting these DL models into practice. This thesis discusses decreasing inference latency through a lightweight 1D Convolutional Neural Network (CNN) model. A dataset containing 55 LoRa devices in an open-set was utilized. Preprocessing methods of Short Time Fourier Transform (STFT) and Fast Fourier Transform (FFT) were compared based on classification accuracy and inference latency. Furthermore, the model that was introduced was evaluated against the 2D CNN model. Even though the lightweight model offers a notable enhancement in inference speed, there are slight and tolerable reductions in its accuracy.
