I was wondering - are there any known techniques to control access to a shared memory object from anywhere but an authorized program?
For instance, lets say I create a shared memory segment for use in a program P, to be accessed by Q, and I make it Read-Write. I can access it using Q because I've given it (Q) the required permissions to do so (running as a particular user with groups, etc).
However, I'm guessing there are instances where someone could potentially access this shared memory from a program R - simply attaching to it and modifying it. To stop this, you could make the memory segment read only - but now program R could still read what was in the memory.
My question is in parts -
Is there a way to,
a) allow only Q to access the shared memory?
b) figure whether a read was done by someone apart from Q - and who did it? [Is this even possible?] For bonus points, could this be done cross-platform? [Probably not, but no harm trying :)]
Under what circumstances could a rogue program attach to the shared memory? I presume one way is if a user is able to exploit OS holes and become the user that started the program. Any others?
POSIX shared memory has the same permissions system as files - if you run
ipcs
you'll see the permissions of the shared memory segments on your system:In answer to question 1a), you can use the normal UNIX permissions system to only allow access from a certain user and/or group. This can be controlled using
shmctl
:For 1b), I don't think any auditing interfaces exist for shared memory access.
With regards to your second question, any process running as the shm owner/group, or running as root will be able to access your memory - this is no different to accessing any other resource. Root can always access anything on a *ix system; and so any exploit which escalated a user to root would allow access to any shared memory region.