基于神经网络的Wi-Fi室内定位安全性问题研究

Research on Wi-Fi indoor location security based on CNN

针对现有的基于深度学习的室内定位解决方案容易受到无线接入点AP攻击的问题,提出了基于卷积神经网络(CNN)的室内定位框架,构建了一种基于深度学习的Wi-Fi指纹室内安全定位系统(DS-LocCNN),使得多建筑和多楼层在面对恶意攻击时的定位精度得到了保证.通过在UJIIndoorLoc数据集上评估此系统,证明了所提出的DS-LocCNN能很好地抵御基于AP的恶意攻击,且在建筑级定位和楼层级定位上的成功率优于现有的解决方案.

Aiming at the problem that the existing indoor positioning solutions based on deep learning are vulnerable to AP(access point) attacks, a positioning framework based on convolutional neural networks(CNN) was proposed, and a deep learning-based Wi-Fi fingerprint indoor security positioning system(DS-LocCNN) was constructed,which enabled multi-building and multi-floor to face malicious attacks.The positioning accuracy during attack was guaranteed.By evaluating the system on the UJIIndoor Loc dataset, it is proved that the DS-LocCNN proposed can well resist AP-based malicious attacks, and its success rate is better than the current solutions in building-level localization and floor-level localization.