In recent years, several stress detection methods have been proposed, usually based on machine learning techniques relying on obstructive sensors, which could be uncomfortable or not suitable in many daily situations. Although studies on emotions are emerging and rising in Software Engineering (SE) research, stress has not been yet well investigated in the SE literature despite its negative impact on user satisfaction and stakeholder performance. In this paper, we investigate whether we can reliably implement a stress detector in a single pipeline suitable for real-time processing following an arousal-based statistical approach. It works with physiological data gathered by the E4-wristband, which registers electrodermal activity (EDA). We have conducted an experiment to analyze the output of our stress detector with regard to the self-reported stress in similar conditions to a quiet office workplace environment when users are exposed to different emotional triggers.