학술논문
A Device Fingerprinting Technique to Authenticate End-user Devices in Wireless Networks
Document Type
Conference
Author
Source
2022 IEEE 2nd International Symposium on Sustainable Energy, Signal Processing and Cyber Security (iSSSC) Sustainable Energy, Signal Processing and Cyber Security (iSSSC), 2022 IEEE 2nd International Symposium on. :1-6 Dec, 2022
Subject
Language
Abstract
Device fingerprinting is a method of identifying a device based on attributes provided by the device configuration and usage. This work utilizes traffic characteristics to generate the fingerprint of end-user devices in a wireless network. In order to create fingerprints, an optimized histogram-driven approach is applied to network traffic. Based on histograms and the density function, the optimized histogram method estimates a bin-width that minimizes expected least square errors (L2). An Extreme Learning Machine (ELM) has been used to classify the devices after fingerprints are generated. The ELM is derived from artificial neural networks, but it is much faster than conventional neural networks. To evaluate the proposed model, two benchmark datasets were used: SIGCOMM-2004 and SIGCOMM-2008. In SIGCOMM-2004, it fingerprinted 74 devices with 96.42% accuracy, while in SIGCOMM-2008, it fingerprinted 48 devices with 86.45% accuracy. Experiments have shown that the proposed method is the most effective.