In this work, a novel nonlinear stress assessment method based on multivariate multiscale entropy has been applied to electroencephalogram (EEG) data collected from subjects who were in resting state and mental counting state. The complexity feature of brain electroencephalogram signals collected during resting and cognitive workload has shown that the subjects who are in mental counting state demonstrate higher complexity compared to subjects who are in resting state. Moreover, the support vector machine (SVM) has shown promising classification between resting and mental counting states by providing 80% sensitivity, 100% specificity and 90% classification accuracy