The threat to IT systems from attacks is continuously increasing. The detection of attacks is very challenging especially for embedded systems as they have very limited resources. In this context we have been investigating successfully lightweight concepts for the detection of network-based attacks, but also for the detection of attacks on individual devices for many years. For network-based attacks the focus is on platform-independent lightweight procedures for the analysis of information on all protocol levels as well as the investigation and realization of AI methods for the detection of attacks, e.g. against critical infrastructures. AI methods are also being investigated and realized as efficiently as possible for the detection of attacks against individual devices, for example in the form of support vector machines for the detection of return oriented programming attacks.