Every AWS account has a root user. It can do anything in the account, including close it, and no IAM policy can restrict it. Too often it was created with one person's email address, a password they remember and an authenticator app on their phone. When that person leaves or loses the phone, you are locked out. If an attacker phishes them, the attacker owns the account.
You cannot remove the management account's root user, so make it hard to misuse and hard to lose.
AWS forces MFA on every root user but accepts any type. Choose a FIDO2 hardware security key such as a YubiKey, because a key will not answer a fake sign-in page the way a six-digit code will. See Enable a passkey or security key for the root user.
Register at least two keys. A root user can have up to 8 MFA devices, and any one is enough to sign in. Give the keys to different custodians, or keep one in the office safe. The spare key means a lost key never becomes an account recovery.
Also:
aws-root@northwind.com. Password resets and security notices go there, and a personal mailbox dies with the person's employmentbob@northwind.com❌ Figure: Bad example - Bad example - The whole organization depends on Bob, his phone and a code that can be phished
aws-root@northwind.com, a group mailbox the infrastructure leads can read✅ Figure: Good example - Good example - A phishing page gets nothing, and losing one person or one key locks nobody out
| Option | Phishing-resistant | Use for root? |
|---|---|---|
| FIDO2 hardware security key | Yes | ✅ Yes |
| Passkey in a shared team vault | Yes | ⚠️ Only if security agrees. Anyone with vault access holds the factor |
| Synced passkey in a personal account (iCloud, Google) | Yes | ❌ No. It follows one person's personal account, not the company |
| Hardware TOTP token | No | ⚠️ Better than a phone, but the code can still be phished |
| Authenticator app (TOTP) | No | ❌ No. One phone, and a code that can be phished |
Member accounts do not need root credentials. In the management account, turn on centralized root access, then delete the root password, access keys, signing certificates and MFA devices from each member account. Nobody can then sign in as root or recover the password, and new accounts start with no root credentials.
An administrator with the sts:AssumeRoot permission can still start a short-lived session scoped to one task: deleting an S3 or SQS queue policy that locks everyone out, or deleting or re-enabling the account's root credentials. Register Log/Security as the delegated administrator for IAM so these sessions do not start from the management account.
A few jobs still need a real root sign-in, such as enabling MFA delete on an S3 bucket. Use Allow password recovery on that one account, reset the password through the root mailbox, do the job, and delete the credentials again. Member accounts therefore need a group mailbox too.
A root sign-in should be rare enough to page someone. See Do you alert on AWS root sign-ins and logging gaps?
Figure: Alert on every root sign-in