This paper analyzes “the vulnerability of a convolutional neural network (CNN)–based indoor localization solution.” The authors “propose a novel methodology to maintain indoor localization accuracy ... in the presence of access point (AP) attacks.”
Indoor localization is an emerging area and the paper is significant in this context. AP attacks can have variable impact on indoor localization accuracy. In extreme cases, such as emergency response, low accuracy in indoor localization can be fatal. The paper elaborates on the threat model and also describes background work.
The experiment section of the paper is well written. The proposed technique, SCNNLOC, is compared with an existing CNN-based indoor localization framework (CNNLOC). On average, SCNNLOC is ten times more secure than CNNLOC. The authors consider various attacks “such as [wireless access point, WAP] spoofing, WAP jamming, and even environmental changes.” An important area that still needs to be explored is time (or processing delay) in estimating the location. For critical applications such as emergency response systems, latency could be critical.