Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
AWS IAM policy evaluation is the fixed order in which identity policies, resource policies, permissions boundaries, SCPs, and session policies combine into a single allow or deny.

How Does AWS IAM Policy Evaluation Work?
Every call to an AWS API is refused unless something explicitly permits it, and one explicit deny anywhere in the chain beats every allow. AWS IAM policy evaluation is the fixed sequence AWS uses to reduce a pile of overlapping documents—identity policies, resource policies, permissions boundaries, service control policies, and session policies—into a single yes or no. Teams that know the order grant exactly what they intended; teams that do not tend to paste in a wildcard and move on.
The starting state of every request is an implicit deny. Something must say Allow for the call to proceed, and nothing may say Deny. That sounds simple until a single principal carries an identity policy from a group, a role-based permissions boundary, an inherited service control policy, and a bucket policy written by another team two years ago.
Two kinds of deny exist, and the difference matters when you debug. An implicit deny is the absence of an allow: add the missing permission and the call works. An explicit deny is a statement with "Effect": "Deny", and no later allow can override it. Explicit denies are the only reliable guardrail, which is why organisation-wide controls are written as denies rather than as carefully trimmed allow lists.
For a request inside one account, AWS evaluates in this sequence and stops at the first deny:
AssumeRole time further narrow the session.The practical rule is that each layer intersects with the others. A permission must survive every applicable boundary to take effect, so adding a policy can never widen access beyond the ceiling set above it.
When a principal in account A touches a resource in account B, both accounts must agree. The identity policy in A has to allow the action, and the resource policy in B has to allow the principal. KMS is the classic trip hazard: a key policy that does not mention the calling role will refuse decryption even though the role holds kms:Decrypt on Resource: "*". S3 adds another twist, because Block Public Access settings are evaluated before the bucket policy and will override a public grant you deliberately wrote.
Condition keys decide whether a statement applies at all. aws:SourceIp, aws:PrincipalOrgID, aws:SourceVpce, and aws:MultiFactorAuthPresent are common, and a mistyped key simply makes the condition never match—so a deny you believed was active quietly does nothing. Use Null and BoolIfExists deliberately, and test rather than assume. Resource-level conditions also interact with least privilege work: narrowing by tag is usually safer than enumerating hundreds of ARNs that drift over time.
Work from the outside in. Check CloudTrail for the failed event and read the error text, which often names the policy type responsible. If the failure came from an SCP, the call is blocked no matter what you change inside the account. The IAM policy simulator reproduces identity policy decisions offline, and IAM Access Analyzer validates policy syntax and flags statements that grant access to external principals. For encoded failures, AWS returns a message that the Security Token Service can decode into the specific denying statement.
A few habits prevent most of the pain: keep the number of policy layers small, prefer tag-based conditions to long resource lists, review permissions boundaries whenever a team asks for a new role, and run IAM Access Analyzer unused-access findings so permissions that nobody exercises are removed rather than inherited forever. Overlapping identity policies also make cloud infrastructure entitlement management tooling worth its licence fee in estates with hundreds of roles.
Policy documents describe intent; only testing shows what an attacker could reach in practice. At Privia Security, we map real privilege paths across AWS accounts as part of our cloud security services.
No. A service control policy only sets the maximum permissions available to accounts in an AWS organisation. Even a policy that allows every action grants nothing on its own; the principal still needs an identity policy or a resource policy that allows the call. This is why removing the default FullAWSAccess policy without adding targeted allows elsewhere breaks an entire account.
Because KMS keys carry their own resource policy. The key policy must name the calling principal, or delegate to IAM with a statement that allows the account root to use IAM policies. If the key policy says nothing about your role, the request is denied regardless of the identity policy. Check the key policy first whenever encryption or decryption fails unexpectedly.
Both are ceilings, but they sit at different levels. A service control policy applies to an entire account or organisational unit and is managed centrally. A permissions boundary is attached to one user or role and is typically used so that a delegated administrator can create roles without being able to grant more than the boundary allows. Many estates use both together.
No. Once any applicable policy contains an explicit deny that matches the request, the decision is final and no allow can reverse it. That makes explicit denies the right tool for non-negotiable controls, such as blocking a region, preventing deletion of audit trails, or stopping changes to logging configuration. Write them carefully, because they also apply to your break-glass roles.
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.