Wireless Device Identification Based on Transfer Learning

The rapid advancements in wireless communication have significantly improved people's lives, offering unprecedented levels of convenience. However, the widespread use of wireless devices has also brought about numerous security threats. IEEE 802.11, a widely adopted wireless LAN standard, is often targeted by attackers who execute security attacks such as address resolution protocol (ARP) poisoning, evil twin attacks, or rogue access points (APs). These attacks manipulate the ARP address table to acquire legitimate users' data or deceive users into connecting to malicious devices, ultimately resulting in the theft of sensitive information. In existing protocols, distinguishing between legitimate and malicious devices is a significant challenge when SSIDs and MAC addresses are identical. To address this issue, we propose a novel approach that identifies various wireless devices by leveraging the hardware's physical fingerprinting based on channel state information and incorporating the concept of transfer learning. Our method effectively extracts features from complex and dynamic wireless environments with a view to enhancing the privacy and security of wireless applications.