ML Comparison: Countermeasure Prediction Using Radio Internal Metrics for BLE Radio

The reliability of low-power wireless communications is being challenged by the proliferation of Internet of Things (IoT) devices. In an increasingly dynamic context, new countermeasure management is needed to make the IoT network more flexible. This paper proposes a comparison between three machine learning (ML) algorithms to predict the next countermeasure to be applied during the next wake-up slot of a BLE receptor. We evaluate the ratio between performance and stability of the solution. The best technique that emerged is the bagged tree, which predicts the future countermeasure to be implemented with an accuracy of 99.7%.