With the rapid advancement of Internet of Things technology, an increasing number of users have adopted wireless cameras for home security. However, malicious actors exploit this trend by deploying hidden cameras indoors for covert surveillance, severely threatening personal privacy. To counter this, several apps detect such devices by analyzing physical characteristics via smartphone cameras or scanning wireless networks. Nonetheless, these solutions suffer from low accuracy and operational complexity, limiting their effectiveness in reliably locating hidden cameras. In this paper, we propose a novel method, CAMLOC, to effectively detect hidden cameras, and it also enables fine-grained classification of camera models and operational states. Moreover, CAMLOC is capable of accurately localizing hidden cameras within the environment. This method is highly agile and generalizable, and it does not require decryption of over-the-air traffic. It identifies camera traffic and determines camera location by analyzing variations in traffic features. The core of CAMLOC consists of four modules: an Air Interface Traffic Collection Module, a Camera and Non-Camera Identification Module based on convolutional neural networks, a Fine-Grained Camera Fingerprint Analysis Module based on gradient boosting decision trees, and a Camera Positioning Module based on Differentiation Method. Experimental results demonstrate that CAMLOC can detect hidden cameras with an accuracy of 99.1%, and classify their models and behaviors with an accuracy of 99.7%, outperforming a wide range of existing approaches. Moreover, it provides effective dynamic camera localization capabilities, which have not been achieved by any prior work.
Zhao et al. (Wed,) studied this question.