On 11/06/2017 10:23 PM, Serge E. Hallyn wrote:

I think I definately prefer what I mentioned in the email to Boris.
Basically a "permanent capability bounding set".  The normal bounding
set gets reset to a full set on every new user_ns creation.  In this
proposal, it would instead be set to the calling task's permanent
capability set, which starts (at boot) full, and which privileged
tasks can pull capabilities out of.

Actually, this may solve a similar problem I've been looking at. The 
idea was basically at strategic points in the kernel (possibly LSM hook 
sites, still evaluating, and probably syscall entry) validate that a 
task has not "magically" acquired capabilities that it or parent 
specifically said it cannot have and then take some action like say 
killing it immediately. Using your terms, basically make the "permanent 
capability set" a write-once privilege escalation defense. To handle the 
0-day threat, perhaps make it writable but only with more "restrictive" 
values.

-chrish