Data-Dependent Fingerprints for Wireless Device Authentication.

摘要

While authenticating wireless radios based on the unique imperfections in their transmitted waveform has become a topic of some interest, such fingerprinting techniques are vulnerable to an attacker who can listen to a radio's transmission and later mimic the transmission using a sophisticated arbitrary waveform generator. However, this type of vulnerability can be reduced if the authentication is accomplished using a random selection from a long list of possible challenge-response pairs and if the node requesting network access has a fingerprint that changes with each valid response. This work provides a first study of such a concept using a tunable filter, used during the authentication exchange, whose tuning voltages are determined from the response data. The work uses simulations and measurements to demonstrate the effectiveness of estimating the distortion function introduced by the tunable filter and using it to identify the device. Results show that the technique can achieve near perfect discrimination between devices and can reject an attacker with very high probability.