AWS Bedrock API Key Phantom User Activity Outside Bedrock

edit
IMPORTANT: This documentation is no longer updated. Refer to Elastic's version policy and the latest documentation.

AWS Bedrock API Key Phantom User Activity Outside Bedrock

edit

Identifies an Amazon Bedrock API key phantom user (an IAM user whose name starts with "BedrockAPIKey-") acting as the caller of a non-Bedrock API request, such as IAM, STS, EC2, VPC, or KMS calls. These users are provisioned by AWS to back a Bedrock bearer token and carry the AmazonBedrockLimitedAccess managed policy, which also grants IAM, VPC, and KMS reconnaissance. A phantom user performing activity outside of Bedrock indicates its credentials are being used beyond their intended scope, which is the privilege-escalation path realized: an attacker who created standard IAM access keys for the phantom user is now using them for reconnaissance or lateral movement outside the Bedrock authentication boundary.

Rule type: query

Rule indices:

  • logs-aws.cloudtrail-*

Severity: high

Risk score: 73

Runs every: 5m

Searches indices from: now-30m (Date Math format, see also Additional look-back time)

Maximum alerts per execution: 100

References:

Tags:

  • Domain: Cloud
  • Domain: LLM
  • Data Source: AWS
  • Data Source: AWS CloudTrail
  • Data Source: Amazon Web Services
  • Data Source: AWS IAM
  • Data Source: Amazon Bedrock
  • Use Case: Threat Detection
  • Tactic: Privilege Escalation
  • Resources: Investigation Guide

Version: 1

Rule authors:

  • Elastic

Rule license: Elastic License v2

Investigation guide

edit

Triage and analysis

Investigating AWS Bedrock API Key Phantom User Activity Outside Bedrock

Amazon Bedrock API key phantom users ("BedrockAPIKey-*") exist only to back a Bedrock bearer token and carry the AmazonBedrockLimitedAccess managed policy. That policy grants Bedrock control-plane actions plus IAM, VPC, and KMS reconnaissance, so if an attacker adds standard IAM access keys (or a console login) to the phantom user, those credentials can be used for reconnaissance and lateral movement well beyond Bedrock.

This rule fires when a "BedrockAPIKey-*" user is the caller of an API request whose service is not Bedrock. Because the phantom user has no legitimate reason to act outside Bedrock, such activity is the privilege-escalation path realized.

Possible investigation steps

  • Review "aws.cloudtrail.user_identity.arn", "event.provider", and "event.action" to understand what non-Bedrock activity the phantom user performed.
  • Inspect "source.ip"/"source.as.number" and "user_agent.original", and determine whether the phantom user holds IAM access keys or a login profile (the escalation pivot).
  • Review the full sequence of the phantom user’s actions for reconnaissance (IAM/EC2/STS enumeration) or attempts to access other resources.
  • Correlate with the credential-addition event (CreateAccessKey/CreateLoginProfile on the same user).

False positive analysis

  • Phantom users are not meant to perform non-Bedrock activity, so this should be rare. Validate any intentional repurposing before excluding the identity.

Response and remediation

  • If unauthorized, disable and remove the phantom user’s IAM access keys and login profile, and delete the phantom user after preserving forensic evidence.
  • Review the account for resources the phantom user may have accessed or modified.
  • Deploy an SCP denying "iam:CreateAccessKey" and "iam:CreateLoginProfile" on "arn:aws:iam::*:user/BedrockAPIKey-*" to prevent the pivot.

Setup

edit

This rule requires AWS CloudTrail logs ingested via the Elastic AWS integration. See https://docs.elastic.co/integrations/aws/cloudtrail for setup details.

Rule query

edit
data_stream.dataset: "aws.cloudtrail"
    and aws.cloudtrail.user_identity.type: "IAMUser"
    and user.name: BedrockAPIKey-*
    and not event.provider: (
        "bedrock.amazonaws.com" or "signin.amazonaws.com" or
        "agreement-marketplace.amazonaws.com" or "discovery-marketplace.amazonaws.com"
    )
    and not event.action: ("GetCallerIdentity" or "GetSessionToken" or "GetAccessKeyInfo")

Framework: MITRE ATT&CKTM