Measures to detect malicious internal messages or activity should be considered (Table B1, 11.1: malicious internal messages such as CAN).
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.