Predictions arising from deep neural networks may be very accurate but not very robust, leading to uncertainty in their outcome. This critical problem is receiving growing attention from the Machine Learning (ML) community. A practical solution that is increasingly applied is calculating confidence bounds for the ML predictions. Most confidence bounds available in the literature are theoretically sound but unfeasible from a practical viewpoint. In this paper, we contribute to the literature with probabilistic confidence bounds based on conditional probabilities, and we demonstrate their operational validity by means of a real-world application that concerns the prediction of the sleeping states of car drivers.