In this talk we present the trust region method (S)NOWPAC, (Stochastic) Nonlinear Optimization With Path-Augmented Constraints, for derivative-free constrained stochastic optimization. We first briefly review the key features, in particular the constraint handling, of the optimizer NOWPAC for deterministic derivative-free optimization. We then generalize the trust region management in NOWPAC to account for noisy evaluations of the objective function and the constraints, and we reduce the negative impact of noise on the quality of surrogate models by using Gaussian process regression. We demonstrate the capabilities and efficiency of (S)NOWPAC by applying it to the robust optimization of a distillation column as well as to several test problems from the CUTEst benchmark set.