hulumi/baseline CVE-2026-82863

Two concrete prerequisites must both be met: first, an attacker must possess AWS IAM credentials with the cloudtrail:PutEventSelectors permission (or equivalent, such as an overly permissive AdministratorAccess role) allowing modification of CloudTrail event selectors in the target account; second, the victim environment must use @hulumi/baseline versions before 1.3.2 as a CloudTrail integrity control and must actually execute the library's checks against the account post-tampering. The detection gap only surfaces when the library's audit run follows a tampering event. Exploitation is blocked entirely for environments without these specific AWS IAM privileges - unauthenticated or low-privilege external attackers cannot trigger the gap independently. Environments using redundant detection mechanisms (AWS Config, Security Hub, third-party CSPM) retain partial coverage even on unpatched versions.

The provided CVSS 4.0 score of 8.7 with AV:N/AC:L/AT:N/PR:N/UI:N/VC:H reflects the maximum potential confidentiality impact of undetected CloudTrail tampering, but overstates the ease of exploitation: PR:N applies to the @hulumi/baseline tool itself, but the underlying attack - modifying AWS CloudTrail event selectors - requires AWS IAM credentials with cloudtrail:PutEventSelectors or equivalent permissions. Real-world risk is therefore conditional on a prior credential compromise or malicious insider scenario. No EPSS score is available in the source data, no CISA KEV listing exists, and no public proof-of-concept exploit has been identified, all of which reduce urgency relative to the raw score. For organizations actively using @hulumi/baseline as a CloudTrail integrity control, the gap is a meaningful blind spot; for those with redundant monitoring via AWS Config, Security Hub, or SIEM integrations, the practical impact is lower. This is a high-priority patch for consumers of this library but not a crisis for the broader ecosystem.

A threat actor with a compromised AWS IAM role that includes CloudTrail modification permissions calls PutEventSelectors to narrow the event selectors and suppress logging of sensitive API calls such as S3 data events or IAM credential operations. When @hulumi/baseline before 1.3.2 next runs its audit checks, the specific selector modification used falls into the library's detection gap, no alert fires, and the attacker continues operating within the AWS environment with a reduced audit trail. No public proof-of-concept exploit for this detection gap has been identified.