Medical devices require the provision of life-critical functionality even under adverse conditions. We imagine to model (at design time) and to switch (at run-time) security modes in a self-adaptive way, thus, reducing attack surfaces in case of a malfunction, attack, or when vulnerabilities become known. Modes return back to normal when patches are provided and installed. Context-aware devices can resiliently provide a degraded mode of operation with a smaller attack surface instead of completely disabling the whole system or a device recall. Healthcare organizations and patients should actively protect themselves by implicitly or explicitly switching to modes with limited activity ranges for attackers. We use simulation to check all circumstances and the self-healing functionality to return to normal mode. In this paper, we present our ongoing work to make medical devices more secure. We discuss how modes can support that, how they are defined, and what challenges they provide.