On Fri, 9 Apr 2021 at 12:19, Stephen John Smoogen <smooge at gmail.com> wrote: > > > On Fri, 9 Apr 2021 at 12:02, Valeri Galtsev <galtsev at kicp.uchicago.edu> > wrote: > >> >> >> On 4/9/21 10:47 AM, Binet, Valere (NIH/NIA/IRP) [C] wrote: >> > The NIST and CIS baselines don't allow su, we have to use sudo on >> government computers. >> > >> >> Could you enlighten me on the rationale behind that restriction? As, as >> you already noticed, my [ancient, maybe] reasoning makes me arrive at an >> opposite conclusion. (but mine is pure security consideration with full >> trust vested into sysadmin, see below...) >> >> On a second guess: it is just for a separation of privileges, and >> accounting of who did what which sudo brings to the table... Right? >> >> > sudo brings into accounting and the ability to restrict a person to a > single command. [That is hard to do well but it is possible.] It also > allows for an easily auditable configuration file set so that you can see > what should have been allowed and what shouldn't. Versus the usual 'oh lets > make it setgid blah or setuid foo but restricted to this group..' and > people forgetting it was done that way or why. > > That said it is like any tool can be used as a hammer when it should have > remained a phillips head. > > Finally sudo can allow for better RBAC rules where if that is needed you had to have multiple su commands that were aligned to each role so that people could not escape their jail. [My understanding is that this is where your chosen OS shines with sudo and this was lifted to other os's laster.] By 2005 most .gov/.mil baselines required su to be no longer allowed because of this. -- Stephen J Smoogen.