MU-MIMO Wi-Fi beamforming feedback matrices database for radio fingerprinting

This database collects the .pcap files acquired by monitoring a private IEEE 802.11ac Wi-Fi network operating in multi-user multi-input multi-output (MU-MIMO) mode. The Wi-Fi network consists of one access point (AP) (beamformer) and two stations (STAs) (beamformees). The AP was implemented through a Gateworks GW6200 single board computer (SBC) equipped with a Compex WLE1216v5-23 IEEE 802.11ac module. Two Netgear Nighthawk X4S AC2600 routers, with 1 or 2 out of 4 antennas enabled, acted as STAs (beamformees). At the AP, 3 antennas were used to sound the channel for downlink (DL) MU-MIMO transmission mode and the STAs were served with 1 or 2 spatial streams each. For the data transmission between the AP and the STAs, we used channel 42, i.e., with carrier frequency of 5.21 GHz and 80 MHz bandwidth. The number of OFDM sub-channels sounded is 234 as the mechanism does not consider the 14 control sub-channels and the 8 pilot ones. The AP uses the quantization parameters 9 and 7 for phi and psi feedback angles, respectively. We generated UDP traffic in the DL direction to induce the AP to trigger the channel sounding mechanism, and collected the angles (phi, psi) that were sent back by the beamformees using the Wireshark network analyzed toolkit running on an off-the-shelf laptop equipped with an IEEE~802.11ac Wi-Fi card. This allows retrieving the beamforming feedback matrices associated with each sounding operation. Two datasets - namely D1 and D2 - were collected. As for the former, the STAs were deployed at different positions to generate different beam patterns and different SNR regimes. The number of enabled antennas is 2 for each beamformer and each of them is served with 2 spatial streams. Dataset D2 was collected while the AP was manually moved in the environment. Here, the number of enabled antennas and spatial streams is 1 for the first beamformee and 2 for the second. The datasets were collected in two different indoor environments. For more information about the setup, please, refer to the related publication. This dataset was used to design and assess the performance of DeepCSI presented in the article ''DeepCSI: Rethinking Wi-Fi Radio Fingerprinting Through MU-MIMO CSI Feedback Deep Learning'' by Francesca Meneghello, Michele Rossi, Francesco Restuccia. The Python source code is available at https://github.com/signetlabdei/DeepCSI. If you use this dataset, please cite our paper: @inproceedings{meneghello2022deepcsi, author = {Meneghello, Francesca and Rossi, Michele and Restuccia, Francesco}, title = {{DeepCSI: Rethinking Wi-Fi Radio FingerprintingThrough MU-MIMO CSI Feedback Deep Learning}}, booktitle = {IEEE International Conference on Distributed Computing Systems}, year = {2022} }

本数据库收录了通过监测运行于多用户多输入多输出（Multi-User Multi-Input Multi-Output, MU-MIMO）模式的私有IEEE 802.11ac无线局域网所获取的.pcap数据包文件。
该无线局域网包含1个接入点（Access Point, AP，即波束成形发射器）与2个站点（Stations, STAs，即波束成形接收器）。AP通过搭载Compex WLE1216v5-23型IEEE 802.11ac模块的Gateworks GW6200单板计算机（Single Board Computer, SBC）实现。两台启用了4根天线中1或2根的Netgear Nighthawk X4S AC2600路由器充当STAs（波束成形接收器）。
在AP侧，共使用3根天线对信道进行探测，以支持下行链路（Downlink, DL）MU-MIMO传输模式，且每个STA分别配置1或2条空间流。AP与STA之间的数据传输采用42号信道，其载波频率为5.21 GHz，带宽为80 MHz。本次探测的正交频分复用（Orthogonal Frequency Division Multiplexing, OFDM）子信道数量为234，该机制未纳入14个控制子信道与8个导频子信道。AP分别针对phi与psi反馈角采用量化参数9和7。
我们在下行链路方向生成用户数据报协议（User Datagram Protocol, UDP）流量，以触发AP的信道探测机制，并通过运行于搭载IEEE 802.11ac无线网卡的商用笔记本电脑上的Wireshark网络分析工具包，采集波束成形接收器回传的角度（phi、psi）数据。由此可获取与每次信道探测操作相关的波束成形反馈矩阵。
本次共采集了两组数据集，分别记为D1与D2。对于D1，我们将STA部署于不同位置，以生成不同的波束成形模式与信噪比（Signal-to-Noise Ratio, SNR）场景。每个波束成形发射器启用2根天线，且每个STA均配置2条空间流。数据集D2的采集过程中，AP在环境中被手动移动，此时第一个波束成形接收器启用1根天线，第二个启用2根天线。两组数据集均采集于两个不同的室内环境中。
如需了解更多实验设置细节，请参阅相关学术论文。本数据集曾用于设计并评估论文《DeepCSI：基于MU-MIMO CSI反馈深度学习的Wi-Fi射频指纹识别新思路》（作者：Francesca Meneghello、Michele Rossi、Francesco Restuccia）中提出的DeepCSI方法的性能。Python源代码可于https://github.com/signetlabdei/DeepCSI 获取。
若您使用本数据集，请引用以下论文：
@inproceedings{meneghello2022deepcsi, author = {Meneghello, Francesca and Rossi, Michele and Restuccia, Francesco}, title = {{DeepCSI: Rethinking Wi-Fi Radio FingerprintingThrough MU-MIMO CSI Feedback Deep Learning}}, booktitle = {IEEE International Conference on Distributed Computing Systems}, year = {2022} }