<![CDATA[This study develops an RFID tag cloning detection system based on edge computing to enhance security in educational environments by integrating three core approaches: Lightweight AES-128 encryption with automatic key rotation every 3 minutes, multidimensional signal fingerprinting analysis (RSSI, phase, and Doppler shift), and ensemble learning-based classification using the k-NN algorithm (k = 5) with dimensionality reduction via PCA. The system is implemented using a Raspberry Pi 4 as the main processor and an Arduino Nano for tag interfacing, supported by an Impinj R420 reader and RTL-SDR v3 for signal acquisition. The dataset consists of 1,200 RFID tag samples (800 genuine and 400 cloned) collected at Universitas Indonesia, with cloned tags generated using Proxmark3 and ChameleonMini. Experimental results show that the system achieves a detection accuracy of 96.7%, validated through 10-fold cross-validation, with a response time of 8.2 ms per tag and power consumption of only 34.6 mW. A Kappa coefficient (κ) of 0.82 indicates high classification consistency. In real-world implementation, the system successfully reduced cloning cases from 15 to just 2 per month. The system outperforms conventional methods such as EPC Gen2, which offers only 78.2% accuracy, and AES-256, which has significantly higher power consumption (62 mW). Its advantages lie in real-time operation with low power consumption, enabled by an edge computing architecture that eliminates the need for an external server. The findings demonstrate the effectiveness of the system as a reliable and practical RFID cloning detection solution suitable for resource-constrained educational environments.]]>
