Stress resilience is defined as a stress coping ability enabling positive adjustment to adverse events. Resilience is recognized as an important occupational factor that influences the performance and health in soldiers, first responders, air traffic controllers and other highly stressful professions. In the context of exposure to potentially traumatic events, the absence of trauma-related psychiatric disorder symptoms is seen as an indicator of resilience. However, in a task-related context resilience is defined as the ability of maintaining normal psychological and physical functioning, when exposed to extraordinary levels of stress and trauma, which is crucial in the context of a highly demanding, safety-critical, and stressful profession such as air traffic control (ATC). Given the above, objectivised assessment of stress resilience has recently emerged as an important research topic. The main problem with psychometric instruments for resilience assessment is their susceptibility to self-report bias, while various genetic, immunological, hormonal or neuroimaging markers of resilience often require expensive and long-lasting assessment procedures. Psychophysiological features of stress resilience are less susceptible to self-report bias than traditional psychological instruments, and could be measured, besides via high-end and highly reliable laboratory equipment, by using various wearable psychophysiological sensors which are nowadays being actively developed and improved. Therefore, the potential of the input-output paradigm for multimodal elicitation and analysis of psychophysiological features of resilience is recognized. This thesis proposes a method for elicitation and computation of psychophysiological features of stress resilience based on the analysis of peripheral physiological signals: electrocardiography (ECG), electromyography (EMG), electrodermal activity (EDA) and respiration; including methodological advances in computation of features related to respiratory sinus arrhythmia (RSA), acoustic startle response (ASR) and psychophysiological allostasis. Experimental validation has shown that the proposed multidimensional resilience feature space reveals differences between an a priori more-resilient and less-resilient group of participants. Furthermore, the predictive power of the obtained feature space is investigated by formulating a binary classification problem: prediction of high- vs. low-performance under stress. Experimental results show that the proposed method for prediction of human performance under stress based on psychophysiological features of resilience and machine learning yields a relatively high classification accuracy. Developed methods and the obtained results are discussed in the context of prior work, while considering limitations and proposing directions for future work.