1. Resources
  2. Detection rules
  3. Lambda Function Opened to Other Accounts or to Anonymous Callers

Lambda Function Opened to Other Accounts or to Anonymous Callers detection rules

An attacker adds a statement to a Lambda function's resource-based policy, or creates a function URL without authentication, so that an outside account or anyone on the internet can invoke the function later.

Platform
AWS
Severity
Medium
Log source
AWS CloudTrail
Last updated
Rule formats
Sigmaand queries for: SplunkMicrosoft SentinelIBM QRadar
What an attacker needs
Credentials for a principal that may add permissions to a function or create or change its function URL.
Impact
An outside account or anyone with the URL can run the function with its execution role until the statement or the URL is removed.
Response
Remove the policy statement and delete the function URL or set it back to AWS_IAM, then contain the identity that made the change.
Mitigation
Let only deployment roles change function policies and URLs, and deny function URLs without authentication with lambda:FunctionUrlAuthType.
CloudTrail events
AddPermission20150331*CreateFunctionUrlConfig*UpdateFunctionUrlConfig*AddLayerVersionPermission*

How the attack works

Every Lambda function has a resource-based policy that says who, apart from the identities of its own account, may use it. AddPermission adds a statement to that policy. The principal can be an AWS service (the normal case, for triggers), another AWS account, an IAM role or user, or a wildcard. A function URL is a second door: a dedicated HTTPS endpoint for the function, created with CreateFunctionUrlConfig, whose AuthType is either AWS_IAM or NONE. With NONE Lambda performs no authentication, and once the resource-based policy grants public access, anyone who has the URL can invoke the function. Layers have a policy of their own: AddLayerVersionPermission shares a layer version with one account, with an organization or with all AWS accounts.

An attacker who has reached an account uses this to keep a way back that does not depend on the compromised identity. The function keeps running with its execution role, so whatever it does for its callers it now does for the attacker: a function that changes IAM, reads secrets or returns data on request is a useful one to open. A shared layer exposes the code and whatever else was packed into it.

Lambda logs these calls as CloudTrail management events. AddPermission appears with the versioned name AddPermission20150331v2, so a rule on the plain API name finds nothing. CreateFunctionUrlConfig, UpdateFunctionUrlConfig and AddLayerVersionPermission are listed without a version suffix. Invocations are data events (Invoke, InvokeFunctionUrl), which CloudTrail does not log by default. AWS publishes no example record for these events: the rules read request fields that the Lambda API reference names, and say so. No public incident report was used for this page.

What an attacker needs

  • Credentials for a principal allowed lambda:AddPermission on the function, or lambda:CreateFunctionUrlConfig or lambda:UpdateFunctionUrlConfig on it.
  • For a layer, lambda:AddLayerVersionPermission on the layer version.
  • A function worth calling from outside, which means one whose execution role or whose output is useful to the attacker.

Impact

An outside account, or anyone with the function URL, can run the function with its execution role for as long as the statement or the URL exists. What that is worth depends on the function: data it returns, actions it performs on input, or simply compute paid for by the owner. The access survives the cleanup of the compromised identity. A shared layer discloses its content to the accounts it was shared with.

Detection rules

Lambda Function Permission Granted to an AWS Account or IAM Principal

Detects Lambda AddPermission calls (logged as AddPermission20150331v2) whose principal is an AWS account ID or an ARN. Most AddPermission calls name an AWS service, because that is how triggers are wired, and those are not matched. A statement for an account, a role or a user gives that identity the right to use the function, and the Lambda API reference says this is how permission is granted to another account. The rule cannot tell whether the account is yours: compare the account ID with your own accounts, and read the Action of the request. The principal is read from requestParameters.principal, which follows the Principal parameter of the Lambda API reference: AWS publishes no example record for this event, so confirm the field in your own events. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.

LevelMediumStatusExperimentalLog sourceAWS CloudTrail
title: Lambda Function Permission Granted to an AWS Account or IAM Principal
id: 0a42ab92-6d61-47bf-81f9-52746904ac1a
status: experimental
description: >-
  Detects Lambda AddPermission calls (logged as AddPermission20150331v2)
  whose principal is an AWS account ID or an ARN. Most AddPermission calls
  name an AWS service, because that is how triggers are wired, and those are
  not matched. A statement for an account, a role or a user gives that
  identity the right to use the function, and the Lambda API reference says
  this is how permission is granted to another account. The rule cannot tell
  whether the account is yours: compare the account ID with your own
  accounts, and read the Action of the request. The principal is read from
  requestParameters.principal, which follows the Principal parameter of the
  Lambda API reference: AWS publishes no example record for this event, so
  confirm the field in your own events.
  It also matches attempts that failed (errorCode present): check errorCode,
  because an unauthorized attempt is worth seeing.
references:
  - https://docs.aws.amazon.com/lambda/latest/dg/logging-using-cloudtrail.html
  - https://docs.aws.amazon.com/lambda/latest/api/API_AddPermission.html
  - https://attack.mitre.org/techniques/T1098/003/
author: Forestall
date: 2026-11-18
tags:
  - attack.persistence
  - attack.privilege-escalation
  - attack.t1098.003
logsource:
  product: aws
  service: cloudtrail
detection:
  selection_event:
    eventSource: lambda.amazonaws.com
    eventName|startswith: AddPermission20150331
  selection_principal:
    requestParameters.principal|re: '^([0-9]{12}|arn:.+)$'
  condition: selection_event and selection_principal
falsepositives:
  - Cross-account integrations inside your own organization, such as a central account that invokes functions in workload accounts, usually deployed by infrastructure-as-code. Add the deployment roles to a filter_* selection on userIdentity.arn.
  - Statements for a role in the same account as the function.
level: medium

False positives

  • Cross-account integrations inside your own organization, such as a central account that invokes functions in workload accounts, usually deployed by infrastructure-as-code. Add the deployment roles to a filter_* selection on userIdentity.arn.
  • Statements for a role in the same account as the function.

Lambda Function Permission Granted to a Wildcard Principal

Detects Lambda AddPermission calls (logged as AddPermission20150331v2) whose principal is the wildcard. Such a statement lets every AWS principal use the function for the action it names, and together with a function URL whose auth type is NONE it lets anyone on the internet invoke the function. AWS's own example of a public function URL policy uses this principal. The rule does not read the conditions of the statement: a wildcard that is limited to one organization with PrincipalOrgID also matches. The principal is read from requestParameters.principal, which follows the Principal parameter of the Lambda API reference: AWS publishes no example record for this event, so confirm the field in your own events. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.

LevelHighStatusExperimentalLog sourceAWS CloudTrail
title: Lambda Function Permission Granted to a Wildcard Principal
id: a4b75788-5643-4559-8f38-21cb5138aafe
status: experimental
description: >-
  Detects Lambda AddPermission calls (logged as AddPermission20150331v2)
  whose principal is the wildcard. Such a statement lets every AWS principal
  use the function for the action it names, and together with a function URL
  whose auth type is NONE it lets anyone on the internet invoke the function.
  AWS's own example of a public function URL policy uses this principal. The
  rule does not read the conditions of the statement: a wildcard that is
  limited to one organization with PrincipalOrgID also matches. The
  principal is read from requestParameters.principal, which follows the
  Principal parameter of the Lambda API reference: AWS publishes no example
  record for this event, so confirm the field in your own events.
  It also matches attempts that failed (errorCode present): check errorCode,
  because an unauthorized attempt is worth seeing.
references:
  - https://docs.aws.amazon.com/lambda/latest/dg/logging-using-cloudtrail.html
  - https://docs.aws.amazon.com/lambda/latest/api/API_AddPermission.html
  - https://docs.aws.amazon.com/lambda/latest/dg/urls-auth.html
  - https://attack.mitre.org/techniques/T1098/003/
author: Forestall
date: 2026-11-18
tags:
  - attack.persistence
  - attack.privilege-escalation
  - attack.t1098.003
logsource:
  product: aws
  service: cloudtrail
detection:
  selection_event:
    eventSource: lambda.amazonaws.com
    eventName|startswith: AddPermission20150331
  selection_wildcard:
    requestParameters.principal|re: '^\*$'
  condition: selection_event and selection_wildcard
falsepositives:
  - Deployments of public web endpoints on function URLs, which need exactly this statement. Add the deployment roles to a filter_* selection on userIdentity.arn and keep alerting on statements added by people or unknown roles.
  - Statements that limit the wildcard to one organization with PrincipalOrgID. Read the request before closing the alert.
level: high

False positives

  • Deployments of public web endpoints on function URLs, which need exactly this statement. Add the deployment roles to a filter_* selection on userIdentity.arn and keep alerting on statements added by people or unknown roles.
  • Statements that limit the wildcard to one organization with PrincipalOrgID. Read the request before closing the alert.

Lambda Function URL Created or Changed Without Authentication

Detects CreateFunctionUrlConfig and UpdateFunctionUrlConfig calls that set the auth type of a function URL to NONE. A function URL is a dedicated HTTPS endpoint for a function, and with NONE Lambda performs no authentication before it invokes the function. The URL can receive requests once the resource-based policy of the function grants public access, which the console adds automatically and which is a separate AddPermission call elsewhere. AWS lists both calls among the events logged for function URLs. The auth type is read from requestParameters.authType, which follows the AuthType parameter of the Lambda API reference: AWS publishes no example record for these events, so confirm the field in your own events. The event names are matched by their start in case a version suffix is present. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.

LevelMediumStatusExperimentalLog sourceAWS CloudTrail
title: Lambda Function URL Created or Changed Without Authentication
id: 15a8d803-059e-4ff7-b275-56e1f7e59575
status: experimental
description: >-
  Detects CreateFunctionUrlConfig and UpdateFunctionUrlConfig calls that set
  the auth type of a function URL to NONE. A function URL is a dedicated
  HTTPS endpoint for a function, and with NONE Lambda performs no
  authentication before it invokes the function. The URL can receive requests
  once the resource-based policy of the function grants public access, which
  the console adds automatically and which is a separate AddPermission call
  elsewhere. AWS lists both calls among the events logged for function URLs.
  The auth type is read from requestParameters.authType, which follows the
  AuthType parameter of the Lambda API reference: AWS publishes no example
  record for these events, so confirm the field in your own events. The
  event names are matched by their start in case a version suffix is
  present.
  It also matches attempts that failed (errorCode present): check errorCode,
  because an unauthorized attempt is worth seeing.
references:
  - https://docs.aws.amazon.com/lambda/latest/dg/urls-monitoring.html
  - https://docs.aws.amazon.com/lambda/latest/api/API_CreateFunctionUrlConfig.html
  - https://docs.aws.amazon.com/lambda/latest/dg/urls-auth.html
  - https://attack.mitre.org/techniques/T1098/003/
author: Forestall
date: 2026-11-18
tags:
  - attack.persistence
  - attack.privilege-escalation
  - attack.t1098.003
logsource:
  product: aws
  service: cloudtrail
detection:
  selection:
    eventSource: lambda.amazonaws.com
    eventName|startswith:
      - CreateFunctionUrlConfig
      - UpdateFunctionUrlConfig
    requestParameters.authType: NONE
  condition: selection
falsepositives:
  - Public web endpoints and webhooks that are built on function URLs on purpose. Add the deployment roles to a filter_* selection on userIdentity.arn and keep alerting on URLs created by people or unknown roles.
level: medium

False positives

  • Public web endpoints and webhooks that are built on function URLs on purpose. Add the deployment roles to a filter_* selection on userIdentity.arn and keep alerting on URLs created by people or unknown roles.

Lambda Layer Version Shared with Other Accounts

Detects AddLayerVersionPermission calls, which add a statement to the resource-based policy of a Lambda layer version. AWS describes the call as the way to grant layer usage permission to other accounts: to one account, to all accounts of an organization, or to all AWS accounts. Every call is therefore a share. Whoever is admitted can download the layer, including code and anything else that was packed into it. Read the Principal and the OrganizationId of the request: a wildcard principal without an organization ID means all AWS accounts. AWS lists the call among the Lambda management events. The event name is matched by its start in case a version suffix is present. It also matches attempts that failed (errorCode present): check errorCode, because an unauthorized attempt is worth seeing.

LevelLowStatusExperimentalLog sourceAWS CloudTrail
title: Lambda Layer Version Shared with Other Accounts
id: 3024e666-b6c6-43e3-b3a5-4bffbab98c49
status: experimental
description: >-
  Detects AddLayerVersionPermission calls, which add a statement to the
  resource-based policy of a Lambda layer version. AWS describes the call as
  the way to grant layer usage permission to other accounts: to one account,
  to all accounts of an organization, or to all AWS accounts. Every call is
  therefore a share. Whoever is admitted can download the layer, including
  code and anything else that was packed into it. Read the Principal and the
  OrganizationId of the request: a wildcard principal without an
  organization ID means all AWS accounts. AWS lists the call among the Lambda
  management events. The event name is matched by its start in case a
  version suffix is present.
  It also matches attempts that failed (errorCode present): check errorCode,
  because an unauthorized attempt is worth seeing.
references:
  - https://docs.aws.amazon.com/lambda/latest/dg/logging-using-cloudtrail.html
  - https://docs.aws.amazon.com/lambda/latest/api/API_AddLayerVersionPermission.html
  - https://attack.mitre.org/techniques/T1098/003/
author: Forestall
date: 2026-11-18
tags:
  - attack.persistence
  - attack.privilege-escalation
  - attack.t1098.003
logsource:
  product: aws
  service: cloudtrail
detection:
  selection:
    eventSource: lambda.amazonaws.com
    eventName|startswith: AddLayerVersionPermission
  condition: selection
falsepositives:
  - Teams that publish shared layers for the accounts of their organization, or vendors that publish public layers on purpose. Add those roles to a filter_* selection on userIdentity.arn.
level: low

False positives

  • Teams that publish shared layers for the accounts of their organization, or vendors that publish public layers on purpose. Add those roles to a filter_* selection on userIdentity.arn.

Detection logic

Patterns that need more than one event, a baseline or knowledge of your own environment. Each is a hypothesis to build and tune in your SIEM, not a ready rule.

Function URL without authentication followed by a public invoke permission

Detection hypothesisSequence of events

A function URL with auth type NONE cannot be called until the resource-based policy grants public access. Outside the console that is a second call. The same principal making both for one function has finished opening it to the internet.

  1. A CreateFunctionUrlConfig or UpdateFunctionUrlConfig event from lambda.amazonaws.com whose request sets AuthType to NONE. Note the function name of the request and the caller in userIdentity.arn.
  2. Within 30 minutes the same caller makes an AddPermission20150331v2 call for the same function whose Principal is the wildcard and whose FunctionUrlAuthType is NONE.
  3. Check what the execution role of the function can do and whether the caller is one of your deployment roles.

EventsCreateFunctionUrlConfigUpdateFunctionUrlConfigAddPermission20150331v2

LimitsEvery public web endpoint built on a function URL is deployed this way, so it is useful only for callers outside the deployment roles. The 30 minutes are a starting point. A URL created in the console has the policy added by Lambda, and how that appears in CloudTrail is not documented.

Function or layer permission names an account that is not yours

Detection hypothesisNeeds your own lists

Granting one outside account the right to invoke a function is an ordinary call with an account ID in it. Whether the account belongs to an attacker or to a partner is not in the event and needs your own account list.

  1. Collect AddPermission20150331v2 and AddLayerVersionPermission events from lambda.amazonaws.com and read the Principal of the request, which is an account ID or an ARN when it is not a service.
  2. Extract the 12-digit account ID and compare it with the accounts of your organization and the third parties you have approved.
  3. Alert when the ID is on neither list, and when the Principal is a wildcard without an organization ID.

EventsAddPermission20150331v2AddLayerVersionPermission

LimitsIt needs a current list of your accounts and approved partners. The Principal comes from the Lambda API reference, and AWS publishes no example record that shows it, so confirm the field in your own events first.

Triage

  • Identify the caller (userIdentity.arn), the source IP address and the user agent, and the function in requestParameters.functionName. Is the caller the pipeline that deploys this function?
  • For AddPermission, read the Principal, Action, SourceArn, SourceAccount and PrincipalOrgID values of the request. An account ID that is not yours, or a wildcard without an organization ID, needs an owner and a reason.
  • For a function URL, read the AuthType of the request, then read the resource-based policy of the function to see whether public invocation is allowed.
  • Check what the execution role of the function can do, and what the function does with its input.
  • Look for a function creation or a code or role change on the same function by the same caller shortly before.

Response

Once the attack is confirmed:

  1. Close the function. Remove the policy statement with RemovePermission, and delete the function URL or set its auth type back to AWS_IAM.
  2. Contain the identity that made the change by deactivating its access keys or revoking its role sessions.
  3. Review what the function did while it was open, using its logs and the CloudTrail activity of its execution role, and revoke the sessions of that role if the function could be made to act for the caller.
  4. Compare the code and configuration of the function with what the pipeline deployed, and redeploy it if they differ.
  5. Check the other functions and layers the identity could change for the same kind of statement or URL.

Mitigation

  • Allow lambda:AddPermission, lambda:CreateFunctionUrlConfig and lambda:UpdateFunctionUrlConfig only to deployment roles, not to workload roles or developers in production accounts.
  • Deny function URLs without authentication where you do not need them, with the condition key lambda:FunctionUrlAuthType in a service control policy or in identity policies.
  • When a statement names an AWS service as principal, set the source ARN or source account, so that resources in other accounts cannot invoke the function through that service.
  • Review public and cross-account access to functions and layers with IAM Access Analyzer.
  • Keep execution roles least-privilege, so that an opened function has little to offer.

Frequently asked questions

Why does a rule that matches the plain name AddPermission find nothing?

Lambda logs this call with a versioned event name, `AddPermission20150331v2`, and AWS lists it that way among the Lambda management events. A rule on the plain API name misses it. The rules on this page match the start of the versioned name, and do the same for the other Lambda events in case a suffix is present.

Does a function URL with auth type NONE mean that anyone can invoke the function?

Not by itself. AWS documents that the resource-based policy of the function is always in effect and must grant public access before the URL can receive requests. The console adds that policy automatically when the URL is created there. With the CLI or the API it is a separate AddPermission call, so look for both events.

Will CloudTrail show the invocations that came from outside?

Only if data events for Lambda are logged. `Invoke` and `InvokeFunctionUrl` are data events, and AWS documents that CloudTrail does not log them by default. Without that setting, use the logs of the function and the activity of the sessions of its execution role.

Has this been seen in real intrusions?

This page was not written from an incident report, and no public report that describes it at the audit event level was used. It follows what AWS documents about resource-based policies, function URLs and layer permissions.

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.