Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Under discretionary access control the owner of a file decides who may read it. Under mandatory access control a system-wide policy decides, and the owner cannot override it.

What Are Mandatory and Discretionary Access Control?
The difference between these two models comes down to a single question: who gets the final say? Under discretionary access control, the owner of a resource does. Under mandatory access control, a policy set by the system does, and the owner’s wishes are irrelevant.
That distinction was formalised in the US Department of Defense’s Trusted Computer System Evaluation Criteria in 1983—the Orange Book—and it still describes the behaviour of the operating systems and cloud platforms you run today.
DAC is what almost everyone experiences. A user creates a file and is its owner; the owner sets permissions on it. Unix mode bits, Windows access control lists, Google Drive sharing, and an S3 bucket policy an engineer edits are all discretionary at heart. The model is flexible, cheap to administer, and matches how people think about their own documents.
Its weakness is structural rather than incidental. Because the owner may grant access to anyone, the system has no way to enforce a rule the owner chooses to ignore. A well-meaning employee shares a folder containing personal data with an external address and the platform complies, because that is exactly what the model promises. Malicious code running as a user inherits the user’s discretion and can re-share anything they own.
This is why most real-world data loss incidents involving cloud storage are not exploits at all. They are permissions working as designed, applied by someone who did not understand the consequences.
MAC removes that latitude. Every subject carries a clearance and every object carries a label, and the kernel or reference monitor compares them against a policy that no ordinary user can change—not even the owner, and often not even root.
The classical formulations come from military confidentiality and commercial integrity requirements. Bell-LaPadula, published in 1973, prevents information flowing downwards: no read up, no write down. Biba inverts it for integrity: no write up, no read down. These sound abstract until you see them implemented.
| Discretionary | Mandatory | |
|---|---|---|
| Who decides | Resource owner | System-wide policy |
| Can the owner override | Yes | No |
| Administrative cost | Low | High |
| Resists malicious or careless users | Poorly | Well |
| Typical use | Files, collaboration, cloud storage | Service confinement, mobile sandboxes, classified systems |
Neither is obsolete and neither is universally correct. Most estates run both: discretionary permissions on the documents people share, mandatory confinement around the services that process them. Layered on top sits role-based access control for organising human permissions, and attribute-based access control where decisions need context.
The first lesson is to stop disabling SELinux. It remains the standard response to a permission denied message in UK server estates, and it converts a confined service into an unconfined one across the whole host. Use audit2allow to read the denial and write a targeted rule, or set the single relevant boolean, rather than running setenforce 0 and moving on.
The second is to recognise where your discretionary controls carry risk the model cannot manage. Anywhere users can share personal data outside the organisation, DAC alone is insufficient under UK GDPR accountability expectations—you need sharing restrictions, sensitivity labels, and data loss prevention enforcing policy above the owner’s discretion. That combination is effectively a mandatory layer imposed on a discretionary platform.
Third, apply confinement to your highest-risk processes first: internet-facing services, file parsers, and anything that handles untrusted input. The value of mandatory controls is highest precisely where exploitation is most likely, and a confined process that an attacker compromises often leaves them with nothing worth having.
At Privia Security, we review host hardening and confinement policy as part of our infrastructure and architecture review services.
Neither, strictly. RBAC is a separate model concerned with organising permissions around job functions, and it can be implemented in a discretionary way where role owners grant membership freely, or in a mandatory way where a central policy controls assignment. In practice most enterprise deployments are discretionary at the edges and centrally governed for privileged roles.
Because a permission denied error from a policy they did not write is faster to silence than to understand, and a great deal of older documentation recommends it. The cost is losing confinement for every service on the host, not just the one that failed. Reading the audit log, generating a targeted rule, or setting the relevant boolean takes minutes and keeps the protection.
It constrains what escalation achieves. An attacker who gains root on a system with enforcing SELinux is still confined by the policy applied to the process they took over, so the files, sockets, and capabilities available may be far narrower than root would normally imply. Policy flaws and unconfined domains remain the way out, which is why policy quality matters.
Object storage permissions are discretionary by default: whoever owns the bucket or object can open it to anyone. Platform-level guardrails such as account-wide public access blocks, organisation policies, and sensitivity labelling act as a mandatory layer on top, enforcing rules that individual owners cannot override. Most large storage exposures happen where that layer is absent.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.