Abstract:
Physical Layer Security relies on detecting suspected behaviors from the communicated device while authenticating it. It makes use of physical layer attributes to secure the communication. Examples of these attributes are the Received Signal Strength (RSS), and Channel State Information (CSI). This work aims to achieve secure communication between Internet of Things (IoT) devices using Artificial Intelligence while relying on some attributes of the physical layer. In particular, the model will be based on time-series measurements from previous and shared states of RSS, for example. Previous work in this domain included the development of a model to predict the location from such measurements. In this work, we will consider these measurements, specifically, time series measurements, for fingerprint authentication. We will develop a model that captures the fingerprint of the transmitter device and authenticate it based on previous measurements and using different types of machine learning algorithms, which will be trained first to compare fingerprints and the training will continue when receiving authenticated fingerprints. Our contribution is mainly in the authentication process where we implement a model on the receiver side that authenticates the communication by authenticating both legitimate parties using a secret key, while accounting for the environmental effects and movements on the communication link.