The problem of security of learning systems and information flows in these systems is a pressing problem in many educational paradigms. In this study, using neural networks, fuzzy logic, system approach and modeling, different approaches to risk management and vulnerability prediction in e-learning process are considered. The main results of the study can be formulated as follows: 1) the levels of training in e-learning are identified and risks are classified by levels (methodological, methodological, modular, didactic); 2) the membership function for the preference of sustainable learning mode at risks is proposed; 3) the filling and use of the knowledge base on risks in the learning system is considered. Examples of implementation of the proposed approaches are given. According to the proposed knowledge base and inference rules it is possible to solve important practical tasks on estimation of potential and state of the learning system. #COMESYSO1120.