Editing Privilege separation (section)

== Implementation ==
A common method to implement privilege separation is to have a computer program [[fork (operating system)|fork]] into two [[process (computing)|process]]es.  The main program drops [[privilege (computing)|privileges]], and the smaller program keeps privileges in order to perform a certain task.  The two halves then communicate via a [[Unix domain socket|socket]] pair.  Thus, any successful attack against the larger program will gain minimal access, even though the pair of programs will be capable of performing privileged operations.

Privilege separation is traditionally accomplished by distinguishing a ''real'' [[user ID]]/[[group ID]] from the ''effective'' user ID/group ID, using the [[setuid]](2)/[[setgid]](2) and related [[system call]]s, which were specified by [[POSIX]]. If these are incorrectly positioned, gaps can allow widespread network penetration.

Many [[computer network|network]] service [[daemon (computer software)|daemons]] have to do a specific privileged operation such as open a [[raw socket]] or an [[Internet socket]] in the [[TCP and UDP port numbers|well known ports]] range. Administrative [[software utility|utilities]] can require particular privileges at [[Run time (program lifecycle phase)|run-time]] as well. Such software tends to separate privileges by revoking them completely after the critical section is done, and change the user it runs under to some unprivileged account after so doing. This action is known as ''dropping root'' under [[Unix-like]] [[operating system]]s. The unprivileged part is usually run under the "[[nobody (username)|nobody]]" user or an equivalent separate user account.

Privilege separation can also be done by splitting functionality of a single program into multiple smaller programs, and then assigning the extended privileges to particular parts using [[file system permissions]]. That way the different programs have to communicate with each other through the operating system, so the scope of the potential vulnerabilities is limited (since a [[crash (computing)|crash]] in the less privileged part cannot be [[exploit (computer security)|exploited]] to gain privileges, merely to cause a [[denial-of-service attack]]).