The probabilistic models and methods for risk prediction and processes optimization, considering characteristics of threats, the measures of control, monitoring and recovery for complex system are proposed. The way of generating new models to increase an accuracy of probabilistic modeling is described. Examples of applications cover detailed studying "human factor", some infrastructure, description of problems for risk analysis and optimization, implementation in the Complex of supporting technogenic safety on the objects of oil & gas distribution. The research is supported by the Russian Scientific Fund.