Pi-Sniffer: Privacy Sniffing through the Wall with WiFi CSI

The popularity and development of smart homes and the Internet of Things also bring the risk of privacy leaks. Illegal sniffing devices outside the target building may receive wireless signals from legitimate signal-sending devices (such as routers, mobile phones, laptops, cameras, etc.), so that the attacker (privacy thief) can steal the privacy information of the target person (victim) without the victim’s awareness. We have designed and implemented Pi-Sniffer, a system that uses off-the-shelf WIFI devices to conduct privacy theft attacks based on wireless sensing, and evaluate the privacy theft ability (accuracy) of attackers under different scenarios. First, we use the RSS signal attenuation model and signal source positioning algorithm to find one or more signal sources located in the target building. And we collect information about the victim by monitoring the signal source or the combination of multiple signal sources in the target building. We then carefully craft features that are independent of the environment. Lastly, we apply LSTM to classify the signals through the wall. Our experiments show that the accuracy of privacy theft attack is affected by the number of active WiFi devices in the victim’s room, and the accuracy of classifying the four types of activity directions of the victim can reach up to 66.7% under different scenarios of monitoring combinations of various active WiFi devices.