Nowadays, occupational stress is a widely experienced phenomenon that manifests itself in our lives consciously or unconsciously. It has always been present at work, however, in recent years it is increasing in intensity and incidence due in large part to new forms of work organization. Prolonged exposure to mental stress contributes to a poor work experience and even to serious health problems. The importance of stress and its significant impact on people's lives underscores the need for early detection. Traditional stress detection solutions, based on standardized questionnaires, come with challenges such as subjective responses and slow implementation. However, advances in the fields of artificial intelligence and affective computing have opened up a wide range of possibilities for early stress detection. Some of these solutions involve the use of data collected by sensors and the application of machine learning techniques for automatic stress detection based on this data. Despite their advantages, some of these solutions face challenges in real-world application, such as intrusiveness and data loss. Therefore, this has motivated the exploration of non-intrusive techniques based on the analysis of behavioural parameters. In this project, priority has been given to the application of detection techniques that do not require additional specific hardware. A widely adopted approach in this domain involves analysing individual behaviour. Specifically, this project seeks to explore how people's interactions with their everyday computers can be exploited to detect their stress levels. To this aim, this project will evaluate the stress predictive performance of various patterns associated with computer interactions, including mouse and keystroke dynamics, eye gaze, and more. Therefore, the main objective will be to develop a stress detection system based on computer interaction using machine learning techniques. To achieve this purpose, several phases have been identified to be followed during the project's development: an exhaustive study of different detection techniques, a previous review and cleaning of datasets to eliminate elements that may interfere in the process, and the application of different machine learning approaches in order to draw conclusions from the collected data. The development of this model will be carried out using the Python programming language.