Device Fingerprinting

Device Fingerprinting for Access Control over a Campus and Isolated NetworkDevice Fingerprinting (DFP) is a technique to identify devices using Inter-Arrival Time (IAT) of packets and without using any other unique identifier. Our experiments include generating graphs of IATs of 100 packets and using Convolutional Neural Network on the generated graphs to identify a device. We did two experiments where the first experiment was on Raspberri Pi and other experiment was on crawdad dataset.First Experiment: Raspberry PiWe developed a packet sniffer application to capture IAT of packets. Packet sniffer application was installed on Raspberry pi that was configured to work as router. We connceted two devices iPad4 and iPhone 7 Plus to the router and created IAT graphs for these two devices. Our scheme based on Convolution Neural Network (CNN) was able to identify the devices with accuracy of 86.7%.DFP on Raspberry PiSecond Experiment: Crawdad DatasetIn the second experiment, we tested the scheme with Crawdad dataset. The proposed scheme achieved accuracy of 95.5% for GTID that is 3% higher than previous scheme \cite{gatech-fingerprinting-20140609} for 14 devices and 5 device types on isolated network while 40% efficient in time to test a device fingerprint.

《面向校园与隔离网络访问控制的设备指纹识别技术》

设备指纹识别（Device Fingerprinting，DFP）是一种无需依赖任何其他唯一标识符，仅通过数据包到达间隔时间（Inter-Arrival Time，IAT）实现设备识别的技术。本实验包含生成100个数据包的到达间隔时间图像，并基于卷积神经网络（Convolutional Neural Network，CNN）对生成的图像开展设备识别任务。我们共开展两组实验：第一组实验基于树莓派（Raspberry Pi）平台，第二组实验基于Crawdad数据集。

第一组实验：树莓派平台
我们开发了一款数据包嗅探应用程序以捕获数据包到达间隔时间。该嗅探程序被部署于配置为路由器的树莓派设备上，将iPad4与iPhone 7 Plus两台设备连接至该路由器，并为这两台设备生成到达间隔时间图像。基于卷积神经网络的识别方案可实现86.7%的设备识别准确率。

树莓派平台设备指纹识别

第二组实验：Crawdad数据集
在第二组实验中，我们使用Crawdad数据集对所提方案进行测试。针对隔离网络中14类设备与5种设备类型的GTID任务，本方案实现了95.5%的准确率，较前人方案cite{gatech-fingerprinting-20140609}提升3%，且设备指纹测试的时间效率提升40%。