Cyber Security News

Google Cloud Composer Flaw Allows Attackers to Gain Elevated Privileges

Research disclosed a now-patched high-severity vulnerability in Google Cloud Platform’s (GCP) Cloud Composer service, dubbed ConfusedComposer.

It could have allowed attackers to hijack cloud workflows and gain control over critical resources. The flaw highlights risks in automated cloud service orchestration.

What Is ConfusedComposer?

Cloud Composer, GCP’s managed Apache Airflow service for workflow automation, relies on Cloud Build (GCP’s CI/CD tool) to install custom Python PyPI packages.

Tenable found that an attacker with the composer.environments.update permission could inject a malicious PyPI package into a Composer environment.

When Cloud Build installs such packages, it automatically runs pre-/post-installation scripts via Python’s Pip tool.

Attackers could exploit this to execute arbitrary code within the Cloud Build instance tied to the default Cloud Build service account—a highly privileged identity with permissions to:

  • Modify Cloud Build pipelines
  • Access Cloud Storage buckets
  • Manipulate Artifact Registry containers
  • Deploy code across projects

Attack Workflow: From Package Upload to Full Compromise

  1. Injection: Attacker adds a malicious PyPI package to a Composer environment.
  2. Execution: Cloud Build installs the package, triggering embedded scripts.
  3. Token Theft: Scripts query GCP’s metadata server to steal the Cloud Build service account token.
  4. Escalation: Attacker uses the token to impersonate the service account, potentially taking over the entire project.
When Cloud Build installs this package in an attempt to build the environment, it uses Pip.

“This bypassed the need for direct access to Cloud Build or Composer service accounts,” said Tenable researcher Gavin Milnthorpe. “The trusted automation pipeline became the attack vector.”

Google resolved the flaw by reconfiguring Composer to use its own service account (with limited permissions) instead of the default Cloud Build account during PyPI installations. Key actions include:

  • Rolling out patches for new Composer environments as of March 2025.
  • Requiring existing instances to update by April 2025.
  • Updating Composer documentation on access control and dependency management.

ConfusedComposer follows 2024’s ConfusedFunction vulnerability, part of a pattern Tenable calls Jenga®-style attacks.

These exploit hidden permissions in cloud services that automatically provision resources (e.g., serverless functions, CI/CD pipelines).

“Cloud providers abstract complexity, but this creates blind spots,” Milnthorpe noted. “Attackers chain these ‘behind-the-scenes’ services to escalate access.”

  • Update Composer: Ensure environments are patched to use the restricted service account.
  • Audit Permissions: Limit composer.environments.update to trusted users.
  • Monitor Cloud Build: Restrict default service account permissions via IAM.

Google confirmed no active exploits were detected before the patch. However, the flaw underscores the need for rigorous oversight in multi-service cloud environments.

Find this News Interesting! Follow us on Google NewsLinkedIn, & X to Get Instant Updates!

Divya

Divya is a Senior Journalist at GBhackers covering Cyber Attacks, Threats, Breaches, Vulnerabilities and other happenings in the cyber world.

Recent Posts

Claude AI Abused in Influence-as-a-Service Operations and Campaigns

Claude AI, developed by Anthropic, has been exploited by malicious actors in a range of…

10 hours ago

Threat Actors Attacking U.S. Citizens Via Social Engineering Attack

As Tax Day on April 15 approaches, a alarming cybersecurity threat has emerged targeting U.S.…

19 hours ago

TerraStealer Strikes: Browser Credential & Sensitive‑Data Heists on the Rise

Insikt Group has uncovered two new malware families, TerraStealerV2 and TerraLogger, attributed to the notorious…

20 hours ago

MintsLoader Malware Uses Sandbox and Virtual Machine Evasion Techniques

MintsLoader, a malicious loader first observed in 2024, has emerged as a formidable tool in…

23 hours ago

Threat Actors Use AiTM Attacks with Reverse Proxies to Bypass MFA

Cybercriminals are intensifying their efforts to undermine multi-factor authentication (MFA) through adversary-in-the-middle (AiTM) attacks, leveraging…

24 hours ago

Threat Actors Target Critical National Infrastructure with New Malware and Tools

A recent investigation by the FortiGuard Incident Response (FGIR) team has uncovered a sophisticated, long-term…

1 day ago