IAM & Access Control

AWS SAA SAA-C03 · Design Secure Architectures

By Abu Anns · Last updated 20 November 2026Exam version: SAA-C03 (current)

What IAM is for

AWS Identity and Access Management (IAM) controls who can do what in an AWS account. Every request to AWS — whether it comes from a person in the console, an application on an EC2 instance, or another AWS service — is authenticated and then authorized by IAM. If the request has no permission that allows it, it is denied. That default-deny behavior is the foundation of AWS security, and the exam tests it constantly: the correct answer is usually the option that grants the smallest permission that still gets the job done.

IAM is a global service, not a Regional one. Users, groups, roles, and policies exist at the account level and apply across all Regions. There is no charge for IAM itself. A few IAM concepts appear in almost every exam question about security, so it pays to learn them precisely rather than approximately.

Identities: users, groups, and roles

An IAM user is a long-term identity for a person or an application. A user signs in with a password (console access) or uses access keys (programmatic access through the CLI, SDKs, or API). Access keys are long-lived credentials: they keep working until they are rotated or deactivated, which is why the exam treats hard-coding them into application code as a mistake. A group is a collection of users; you attach policies to the group and every member inherits them. Groups make permission management scalable — when someone joins the database team, you add them to the group instead of editing their individual permissions.

An IAM role is an identity with permissions but no long-term credentials of its own. Something assumes a role — a user, an application, an AWS service, or an identity from another account — and receives temporary security credentials that expire. This is the single most important IAM idea for the exam: roles, not users with access keys, are how applications and services should get permissions. An EC2 instance that needs to read from S3 gets an instance role (an instance profile); a Lambda function gets an execution role; a developer in another account gets a cross-account role. Temporary credentials from the Security Token Service (STS) limit the damage if credentials leak, because they stop working within hours.

Policies: how permissions are written

A policy is a JSON document that states which actions are allowed or denied on which resources, optionally under conditions. Identity-based policies attach to users, groups, or roles. Resource-based policies attach to the resource itself — an S3 bucket policy or a KMS key policy, for example — and are the only way to grant access to a principal in another account without that account creating a role. Inline policies live inside one identity; AWS managed policies are maintained by AWS and cover common job functions (such as ReadOnlyAccess or AmazonS3FullAccess); customer managed policies are ones you write yourself and can reuse.

Policy evaluation follows a strict order that the exam loves to test. Every request starts as an implicit deny. If any applicable policy allows the action, the request is allowed — unless any applicable policy contains an explicit deny, which overrides every allow. Conditions refine policies further: a policy can require that requests come from a specific source IP, use HTTPS, or include multi-factor authentication (MFA). The condition key aws:MultiFactorAuthPresent is the standard way to require MFA before sensitive actions are permitted.

Least privilege is the design principle behind all of this: grant only the permissions a task actually needs, for only as long as it needs them. AWS supports this with tools like IAM Access Analyzer, which examines access activity and suggests policies scoped to what is actually used, and with permission boundaries, which set a maximum permission ceiling for an identity even if other policies grant more. Service control policies (SCPs) from AWS Organizations work at a higher level still: they limit what is possible in an entire account, regardless of what IAM policies inside that account allow. Note that SCPs do not grant permissions — they can only restrict them.

Protecting the root user and human access

Every AWS account has a root user with unrestricted access to everything, including billing and account settings that IAM cannot control. The exam expects you to know that the root user should be locked down: enable MFA on it, delete its access keys, use it only for the handful of tasks that require it (such as changing the support plan or closing the account), and never use it for daily work. For everyday human access, the modern answer is IAM Identity Center (the successor to AWS SSO), which gives workforce users single sign-on across accounts, ideally federated from an existing corporate identity provider. Creating one IAM user per employee in every account is the old pattern and is hard to manage at scale.

For applications and external identities, the pattern is the same idea: avoid long-term credentials. Web and mobile applications authenticate users through Amazon Cognito, which hands out temporary AWS credentials for authenticated (or guest) users. Workloads running outside AWS can use IAM Roles Anywhere or OIDC federation instead of storing access keys. And whenever you need to know who has access to what, IAM provides credential reports, access advisor data showing when a permission was last used, and Access Analyzer findings that flag resources shared outside the account.

Exam mindset for IAM

IAM questions usually describe a requirement and ask for the most secure or least-privilege way to meet it. The answer is almost never “create an IAM user with AdministratorAccess and share the keys.” Look for the option that uses a role, a managed policy scoped to the job, a condition, or federation. When two options both work, the one with narrower permissions or shorter-lived credentials is usually the better design. And when a question involves another account, remember the two halves of cross-account access: the trusting account's role must have a trust policy naming the other account, and the trusted account's identity must have permission to assume the role.

Permission boundaries, SCPs, and keeping access tidy over time

As an account grows, two more IAM mechanisms help keep permissions under control. A permissions boundary is a managed policy that sets the maximum permissions an identity can have: even if other policies grant more, the boundary caps what is actually allowed. Boundaries are commonly used to let teams create their own roles safely — a developer can build a role for a new application, but the boundary prevents them from granting it administrator access. Service control policies (SCPs) from AWS Organizations work one level higher: they limit what is possible in an entire account or organizational unit, regardless of what IAM policies inside that account allow. An SCP can, for example, prevent anyone in a member account from disabling CloudTrail or from launching resources in unapproved Regions. Remember the key rule: SCPs and boundaries can only restrict permissions, never grant them.

Access also needs maintenance. The IAM credential report shows the state of every user and credential in the account — password age, access key age, MFA status — and is the first thing to check in an audit. IAM Access Analyzer goes further: it examines actual access activity and identifies unused permissions, suggests least-privilege policies based on what was really used, and flags resources shared outside the account. Access keys should be rotated regularly, and unused users and roles removed. The exam rewards designs where access is granted through groups and roles, reviewed periodically, and removed promptly when no longer needed.

Advertisement

Exam traps to watch for

Practice questions

Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.

Q1.A company runs a web application on Amazon EC2 instances. The application needs to read objects from a private Amazon S3 bucket. A developer suggests creating an IAM user with an access key and storing the key in the application configuration file. What should the solutions architect recommend instead?

Report an error

Q2.A new developer joins a team that manages a production database. The developer needs to view database performance metrics in the AWS Management Console but must not be able to modify or delete any resources. Which approach follows the principle of least privilege?

Report an error

Q3.Company A needs to give a partner company read access to objects in one of its S3 buckets. The partner's employees already have IAM users in the partner's own AWS account. What is the most secure way to grant this access?

Report an error

Q4.A company requires that any IAM user who performs sensitive actions, such as deleting an S3 bucket, must have authenticated with multi-factor authentication (MFA) in the current session. Which solution enforces this requirement?

Report an error

Q5.A solutions architect is reviewing the security of a new AWS account. Which action is the MOST important first step for protecting the account's root user?

Report an error

Q6.An IAM user has the AmazonS3FullAccess managed policy attached. A separate policy attached to the same user explicitly denies the s3:DeleteBucket action. The user attempts to delete an S3 bucket. What happens?

Report an error

Q7.A company with 200 employees uses five AWS accounts. Employees currently have a separate IAM user in each account, and the security team struggles to remove access promptly when people leave. Which solution BEST simplifies workforce access management?

Report an error

Q8.A mobile application needs to let authenticated users upload photos directly to an S3 bucket. The application must not embed any AWS credentials in the mobile app. Which solution meets these requirements?

Report an error