In information security, computer science, and other fields, the principle of least privilege, also known as the principle of minimal privilege or just least privilege, requires that in a particular abstraction layer of a computing environment, every module (such as a process, a user or a program depending on the subject) must be able to access only the information and resources that are necessary for its legitimate purpose.[1][2]
In other words, this means giving a user only those privileges which are essential to do his/her work. For example, a backup user does not need to install software, hence the backup user has rights only to run backup and backup-related applications. Any other privileges like installing software etc. are blocked. The principle applies also to a single home PC user who usually does work in a normal user account, and opens their privileged, password protected account (that is, a superuser) only when the situation absolutely demands it.
When applied to users, the terms least user access or least-privileged user account (LUA) are also used, referring to the concept that all users at all times should run with as few privileges as possible, and also launch applications with as few privileges as possible. LUA bugs occur when applications do not work correctly without elevated privileges.
Contents |
The principle of least privilege is widely recognized as an important design consideration in enhancing the protection of data and functionality from faults (fault tolerance) and malicious behavior (computer security).
The principle of least privilege is also known as the principle of least authority (POLA).
The kernel always runs with maximum privileges since it is the operating system core and has hardware access. One of the principal responsibilities of an operating system, particularly a multi-user operating system, is management of the hardware's availability and requests to access it from running processes. When the kernel crashes, the mechanisms by which it maintains state also fail. Even if there is a way for the CPU to recover without a hard reset, the code that resumes execution is not always what it should be. Security continues to be enforced, but the operating system can't respond to the failure properly because detection of the failure wasn't possible. This is because kernel execution either halted or the program counter resumed execution from somewhere in endless, and — usually — non-functional loop.
If execution picks up, after the crash, by loading and running trojan code, the author of the trojan code can usurp control of all processes. The principle of least privilege forces code to run with the lowest privilege/permission level possible so that, in the event this occurs — or even if execution picks up from an unexpected location — what resumes execution does not have the ability to do bad things. One method used to accomplish this can be implemented in the microprocessor hardware. In x86 architecture, the manufacturer designed four (ring 0 - ring 3) running "modes". (This term can be confusing because the term "mode" is used in certain OS variants to refer cumulatively, to the state of the set of bits associated with a given resource).
Least privilege is widely misunderstood and, in particular, is almost always confused with the Trusted Computer System Evaluation Criteria (TCSEC) concept of trusted computing base (TCB) minimization. Minimization is a far more stringent requirement that is only applicable to the functionally strongest assurance classes, viz., B3 and A1 (which are evidentiarily different but functionally identical).
Least privilege is often associated with privilege bracketing, that is, assuming necessary privileges at the last possible moment and dismissing them as soon as no longer strictly necessary, therefore ostensibly avoiding fallout from erroneous code that unintentionally exploits more privilege than is merited. Least privilege has also been interpreted in the context of distribution of discretionary access control (DAC) permissions, for example asserting that giving user U read/write access to file F violates least privilege if U can complete his authorized tasks with only read permission.
As implemented in some operating systems, processes execute with a potential privilege set and an active privilege set. Such privilege sets are inherited from the parent as determined by the semantics of fork(). An executable file that performs a privileged function—thereby technically constituting a component of the TCB, and concomitantly termed a trusted program or trusted process—may also be marked with a set of privileges, a logical extension of the notions of set user ID and set group ID. The inheritance of file privileges by a process are determined by the semantics of the exec() family of system calls. The precise manner in which potential process privileges, actual process privileges, and file privileges interact can become complex. In practice, least privilege is practiced by forcing a process to run with only those privileges required by the task. Adherence to this model is quite complex as well as error-prone.
Historically, the oldest instance of least privilege is probably the source code of login.c, which begins execution with super-user permissions and—the instant they are no longer necessary—dismisses them via setuid() with a non-zero argument.
In practice, true least privilege is neither definable nor possible to enforce. Currently, there is no method that allows evaluation of a process to define the least amount of privileges it will need to perform its function. This is because it is not possible to know all the values of variables it may process, addresses it will need, or the precise time such things will be required. Currently, the closest practical approach is to eliminate privileges that can be manually evaluated as unnecessary. The resulting set of privileges still exceeds the true minimum required privileges for the process.
Another limitation is the granularity of control that the operating environment has over privileges for an individual process.[4] In practice, it is rarely possible to control a process' access to memory, processing time, I/O device addresses or modes with the precision needed to facilitate only the precise set of privileges a process will require.
The original formulation is from Jerome Saltzer:
Every program and every privileged user of the system should operate using the least amount of privilege necessary to complete the job. (Protection and the Control of Information Sharing in Multics, CACM 1974, volume 17, issue 7, page 389)
Peter J. Denning, in his paper "Fault Tolerant Operating Systems" set it in a broader perspective among four fundamental principles of fault tolerance.
Dynamic assignments of privileges was earlier discussed by Roger Needham in 1972.[5][6]
|
||||||||||||||||||||