1. Resources
  2. Detection rules
  3. Secrets Retrieved in Bulk From AWS Secrets Manager

Secrets Retrieved in Bulk From AWS Secrets Manager detection rules

An attacker who controls an identity with broad secretsmanager:GetSecretValue reads many secrets at once, or attaches a resource policy that lets another principal read them, and collects database passwords, API keys and tokens.

Platform
AWS
Severity
High
Log source
AWS CloudTrail
Last updated
Rule formats
Sigmaand queries for: SplunkMicrosoft SentinelIBM QRadar
What an attacker needs
Credentials for a principal allowed secretsmanager:GetSecretValue on many secrets, usually through a wildcard resource or action.
Impact
Working credentials for whatever the secrets protect, such as databases, third-party services and other cloud accounts, which stay valid until each secret is rotated.
Response
Revoke the sessions or deactivate the access keys of the identity that made the calls, then rotate every secret that identity could read.
Mitigation
Allow secretsmanager:GetSecretValue only on the secret ARNs each workload needs, and remove statements that allow it on every resource or through wildcard actions.
CloudTrail events
GetSecretValuePutResourcePolicyBatchGetSecretValue

How the attack works

AWS Secrets Manager holds the credentials that workloads fetch at run time: database passwords, API keys, OAuth tokens. A principal allowed secretsmanager:GetSecretValue on a secret receives its decrypted value. BatchGetSecretValue returns up to 20 secrets in one call, chosen by name or by a filter. An identity policy that allows these actions on every resource, or a wildcard such as secretsmanager:*, makes every secret in the account readable by whoever holds that identity.

Attackers who take over such an identity (a workload role, a pipeline role, an administrator) list the secrets and read them. What they collect are credentials for systems outside IAM, such as databases, SaaS platforms and other clouds, so the compromise spreads beyond the AWS account and outlasts the original foothold. A quieter variant attaches a resource policy to a secret (PutResourcePolicy), so that another principal, possibly in another account, can keep reading it later.

Secrets Manager logs all of its API calls as CloudTrail management events under secretsmanager.amazonaws.com. A batch call writes one BatchGetSecretValue entry and a GetSecretValue entry for each secret requested. Single GetSecretValue calls are what every application makes all day, so a rule on that event alone is noise. The batch call and a resource policy change are far less common, and that is what the rules here match. Systems Manager Parameter Store keeps the same kind of data in SecureString parameters; these rules do not cover it. An attacker who reads secrets one at a time still stands out by breadth: a threshold rule counts how many different secrets one principal requests with GetSecretValue in a short window.

What an attacker needs

  • Credentials for a principal allowed secretsmanager:GetSecretValue on many secrets, usually through a wildcard resource or action. The batch call also needs secretsmanager:BatchGetSecretValue, and secretsmanager:ListSecrets when it selects secrets by filter.
  • kms:Decrypt on the key, for secrets encrypted with a customer managed key instead of the AWS managed key aws/secretsmanager.
  • For the resource policy variant, secretsmanager:PutResourcePolicy on the secret. Reading from another account also requires the secret to use a customer managed key whose key policy admits that account.

Impact

The attacker holds working credentials for whatever the secrets protect: databases, third-party services, pipelines, other cloud accounts. Those credentials are used outside AWS, where CloudTrail sees nothing, and they stay valid until each secret is rotated. A resource policy that admits an outside principal keeps the secret readable after the compromised identity has been cleaned up.

Detection rules

Many Different Secrets Retrieved by One Principal

Counts, per calling principal (userIdentity.arn), how many different secrets it requested from AWS Secrets Manager with GetSecretValue. An application reads the few secrets it was built for, again and again. Ten or more different secrets within 10 minutes is the pattern of someone who lists the secrets of an account and reads them one after another, which the rule on the batch call does not see. AWS documents that Secrets Manager writes CloudTrail entries for all of its operations. The secret is read from requestParameters.secretId, which follows the SecretId request parameter of the Secrets Manager API reference and is the field AWS shows in its example records of other Secrets Manager operations. It holds what the caller sent, a name or an ARN, so one secret that is requested in both forms is counted twice. Denied attempts are counted as well. The count and the window are starting points to tune. For role sessions the principal is the session ARN, so one role used with different session names is counted per session.

Fires when one userIdentity.arn has at least 10 different requestParameters.secretId values within 10 minutes.

Threshold ruleLevelMediumStatusExperimentalLog sourceAWS CloudTrail
title: Secret Value Retrieved (Many Secrets Base Event)
id: b4ab407f-494b-47b4-8784-4aa6789e87f4
name: aws_secrets_manager_secret_value_retrieved
status: experimental
description: >-
  Base event of the threshold rule in this file. Any GetSecretValue call to
  AWS Secrets Manager, including attempts that were denied.
logsource:
  product: aws
  service: cloudtrail
detection:
  selection:
    eventSource: secretsmanager.amazonaws.com
    eventName: GetSecretValue
  condition: selection
level: low
---
title: Many Different Secrets Retrieved by One Principal
id: e8a8fcb4-e74e-443b-b953-8b1e5f3d8f73
status: experimental
description: >-
  Counts, per calling principal (userIdentity.arn), how many different
  secrets it requested from AWS Secrets Manager with GetSecretValue. An
  application reads the few secrets it was built for, again and again. Ten
  or more different secrets within 10 minutes is the pattern of someone who
  lists the secrets of an account and reads them one after another, which
  the rule on the batch call does not see. AWS documents that Secrets
  Manager writes CloudTrail entries for all of its operations. The secret
  is read from requestParameters.secretId, which follows the SecretId
  request parameter of the Secrets Manager API reference and is the field
  AWS shows in its example records of other Secrets Manager operations. It
  holds what the caller sent, a name or an ARN, so one secret that is
  requested in both forms is counted twice. Denied attempts are counted as
  well. The count and the window are starting points to tune. For role
  sessions the principal is the session ARN, so one role used with
  different session names is counted per session.
references:
  - https://docs.aws.amazon.com/secretsmanager/latest/userguide/cloudtrail_log_entries.html
  - https://docs.aws.amazon.com/secretsmanager/latest/apireference/API_GetSecretValue.html
  - https://repost.aws/knowledge-center/secrets-manager-arn
  - https://attack.mitre.org/techniques/T1555/006/
author: Forestall
date: 2026-11-18
tags:
  - attack.credential-access
  - attack.t1555.006
correlation:
  type: value_count
  rules:
    - aws_secrets_manager_secret_value_retrieved
  group-by:
    - userIdentity.arn
  timespan: 10m
  condition:
    field: requestParameters.secretId
    gte: 10
falsepositives:
  - Secret rotation jobs, backup and inventory tools, and platform services that load the secrets of many applications, such as a secrets operator in a cluster. Exclude the roles they use.
  - Deployments that start many services at once under one role. Give such roles a higher count.
  - The values of 10 different secrets and 10 minutes are starting points, not tested limits. Measure what your own workloads produce and adjust both.
level: medium

False positives

  • Secret rotation jobs, backup and inventory tools, and platform services that load the secrets of many applications, such as a secrets operator in a cluster. Exclude the roles they use.
  • Deployments that start many services at once under one role. Give such roles a higher count.
  • The values of 10 different secrets and 10 minutes are starting points, not tested limits. Measure what your own workloads produce and adjust both.

Resource Policy Attached to an AWS Secrets Manager Secret

Detects PutResourcePolicy calls to Secrets Manager, which attach or replace the resource policy of a secret. A resource policy can let another principal, including one in another account, read the secret value, so it is a way to keep access to a secret after the access of the calling identity is removed. The rule cannot tell who the policy admits: read the policy in the request parameters. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.

LevelMediumStatusExperimentalLog sourceAWS CloudTrail
title: Resource Policy Attached to an AWS Secrets Manager Secret
id: ade84f0e-e7a2-4ed6-b27a-954fd7d37163
status: experimental
description: >-
  Detects PutResourcePolicy calls to Secrets Manager, which attach or replace
  the resource policy of a secret. A resource policy can let another principal,
  including one in another account, read the secret value, so it is a way to
  keep access to a secret after the access of the calling identity is removed.
  The rule cannot tell who the policy admits: read the policy in the request
  parameters.
  It also matches attempts that failed (errorCode present): check errorCode,
  because an unauthorized attempt is worth seeing.
references:
  - https://docs.aws.amazon.com/secretsmanager/latest/userguide/cloudtrail_log_entries.html
  - https://docs.aws.amazon.com/secretsmanager/latest/apireference/API_PutResourcePolicy.html
  - https://docs.aws.amazon.com/secretsmanager/latest/userguide/auth-and-access_examples_cross.html
  - https://attack.mitre.org/techniques/T1555/006/
author: Forestall
date: 2026-11-14
tags:
  - attack.credential-access
  - attack.t1555.006
logsource:
  product: aws
  service: cloudtrail
detection:
  selection:
    eventSource: secretsmanager.amazonaws.com
    eventName: PutResourcePolicy
  condition: selection
falsepositives:
  - Infrastructure-as-code pipelines and platform teams that set resource policies when a secret is created, or share a secret with another account on purpose. Add the deployment roles to a filter_* selection on userIdentity.arn and review calls made by anyone else.
  - Policies that only restrict access (deny statements) also match.
level: medium

False positives

  • Infrastructure-as-code pipelines and platform teams that set resource policies when a secret is created, or share a secret with another account on purpose. Add the deployment roles to a filter_* selection on userIdentity.arn and review calls made by anyone else.
  • Policies that only restrict access (deny statements) also match.

Secrets Retrieved in a Batch From AWS Secrets Manager

Detects BatchGetSecretValue calls, which return the values of up to 20 Secrets Manager secrets in one request, selected by name or by filter. Applications usually read one secret at a time with GetSecretValue, so a batch call from an identity that does not normally make one deserves a look. Secrets Manager also writes a GetSecretValue entry for each secret requested; use those to see which secrets were involved. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.

LevelMediumStatusExperimentalLog sourceAWS CloudTrail
title: Secrets Retrieved in a Batch From AWS Secrets Manager
id: b6dd27c9-e5f4-4cbd-a365-42f54faf2e17
status: experimental
description: >-
  Detects BatchGetSecretValue calls, which return the values of up to 20
  Secrets Manager secrets in one request, selected by name or by filter.
  Applications usually read one secret at a time with GetSecretValue, so a
  batch call from an identity that does not normally make one deserves a look.
  Secrets Manager also writes a GetSecretValue entry for each secret requested;
  use those to see which secrets were involved.
  It also matches attempts that failed (errorCode present): check errorCode,
  because an unauthorized attempt is worth seeing.
references:
  - https://docs.aws.amazon.com/secretsmanager/latest/userguide/cloudtrail_log_entries.html
  - https://docs.aws.amazon.com/secretsmanager/latest/apireference/API_BatchGetSecretValue.html
  - https://docs.aws.amazon.com/guardduty/latest/ug/guardduty_finding-types-iam.html
  - https://attack.mitre.org/techniques/T1555/006/
author: Forestall
date: 2026-11-14
tags:
  - attack.credential-access
  - attack.t1555.006
logsource:
  product: aws
  service: cloudtrail
detection:
  selection:
    eventSource: secretsmanager.amazonaws.com
    eventName: BatchGetSecretValue
  condition: selection
falsepositives:
  - Applications, deployment tools and secret-loading components that fetch several secrets in one call. They do it on every start or refresh, from the same role; add those roles to a filter_* selection on userIdentity.arn once confirmed.
  - Administrators and audit scripts that export or compare secrets. Confirm the change record; a person reading many production secrets is rarely routine.
level: medium

False positives

  • Applications, deployment tools and secret-loading components that fetch several secrets in one call. They do it on every start or refresh, from the same role; add those roles to a filter_* selection on userIdentity.arn once confirmed.
  • Administrators and audit scripts that export or compare secrets. Confirm the change record; a person reading many production secrets is rarely routine.

Triage

  • Identify the caller (userIdentity.arn), the source IP address and the user agent. Is this a workload that normally loads secrets in a batch, calling from where it normally runs?
  • Find the GetSecretValue entries written for the same caller at the same time; they name the secrets that were requested.
  • Look for ListSecrets calls and for denied GetSecretValue calls (errorCode present) by the same principal shortly before. Listing everything and probing what is readable is collection, not application behaviour.
  • For PutResourcePolicy, read the policy in the request parameters. A principal from another account, or a wildcard principal, needs an owner and a reason.
  • If the activity is not explained, rotate every secret the principal could read, revoke its sessions and review what the stolen credentials give access to.

Response

Once the attack is confirmed:

  1. Stop the reading. Revoke the active sessions of the role that made the calls, or deactivate the access keys if it was an IAM user, and narrow its policy before the identity is used again.
  2. Remove resource policy statements that admit a principal you do not recognize, on every secret the caller could change and not only on the one in the alert.
  3. Rotate the secrets the principal could read, starting with those named in the GetSecretValue entries and with credentials for systems that can be reached from the internet. A stolen value stays valid until its secret is rotated.
  4. Review the logs of the systems the secrets protect, such as databases, SaaS platforms and other cloud accounts, for use of the stolen credentials. CloudTrail does not see that activity.
  5. Afterwards, confirm that no secret has cross-account access you did not intend and that the identity can read only the secrets it needs.

Mitigation

  • Allow secretsmanager:GetSecretValue only on the secret ARNs each workload needs; remove statements that allow it on every resource and wildcard actions such as secretsmanager:*.
  • Allow secretsmanager:BatchGetSecretValue only to the workloads that really load several secrets in one call.
  • Encrypt sensitive secrets with a customer managed KMS key and restrict its key policy, so that an over-broad IAM policy alone is not enough to read them.
  • Require the condition key secretsmanager:BlockPublicPolicy in every policy that allows secretsmanager:PutResourcePolicy, and review cross-account access to secrets with IAM Access Analyzer.
  • Turn on automatic rotation, so that a stolen value stops working without anyone having to notice the theft.

Frequently asked questions

Why is there no rule on GetSecretValue itself?

Every application that uses Secrets Manager calls `GetSecretValue` all day, so a rule on that event alone is noise. The batch call and a resource policy change are far less common, which makes them usable as alerts. An attacker who reads secrets one at a time is matched only by the threshold rule, which counts the different secrets one principal requests in a short window.

How do I see which secrets a batch call asked for?

A batch call writes one `BatchGetSecretValue` entry and a `GetSecretValue` entry for each secret requested. Search for the `GetSecretValue` entries of the same caller at the same time to get the list. One batch call can return up to 20 secrets, chosen by name or by a filter.

Do these rules cover Systems Manager Parameter Store?

No. Parameter Store keeps the same kind of data in SecureString parameters, but the rules on this page match only events of `secretsmanager.amazonaws.com`. Reads of parameters need their own detection.

Does the resource policy rule tell me who was given access to the secret?

No. It fires on every `PutResourcePolicy` call to Secrets Manager, including policies that only restrict access. Read the policy in the request parameters. A principal from another account, or a wildcard principal, needs an owner and a reason.

References

We respect your privacy

We use cookies to keep this site secure and working properly. With your permission, we also use optional cookies to understand usage and improve the experience. Cookie Policy

You can change your choice at any time.