Barracuda Confirms 44-Hour Cloud Breach Tied to AI Vendor Key and ARTEX Framework

Barracuda says an attacker used a cloud key leaked from an AI vendor's GitHub repository to roam 10 cloud accounts in 44 hours, reaching XDR integration credentials.
Official Barracuda logo with blue fins and white wordmark centered on a black background with neon orange halftone chevrons in opposite corners and a faint gray cloud server rack silhouette

Barracuda Networks has confirmed that an attacker spent roughly 44 hours moving through parts of its internal cloud environment in early November and reached a system storing credentials Barracuda XDR uses to connect to customers’ Microsoft 365, Google Workspace and endpoint security platforms. The company said the intrusion began with a cloud access key it had issued to a third-party AI vendor and never retired, and that the attacker’s behavior matches ARTEX, the open-source AI-driven penetration-testing framework now the subject of an FBI warning.

Barracuda disclosed the incident in a Trust Center advisory published November 10 and updated November 13. Chief Information Security Officer Arve Kjoelen followed with a technical write-up, “Forty-four hours: Anatomy of a machine-speed intrusion”.

A vendor key that was never retired

According to Kjoelen, Barracuda gave the access key to an AI company for an earlier project and did not revoke it when the work ended. “That is on us,” he wrote. In October, the vendor’s own development environment was compromised. Barracuda’s November 13 update confirmed the key had been stored in a configuration file inside the vendor’s GitHub repository, which was exposed in that compromise. Other Barracuda credentials in the same file had already expired or have since been revoked.

Barracuda has not named the vendor. The vendor contacted Barracuda on November 4 about its own incident and shared what it knew.

From one key to 10 cloud accounts

Barracuda’s timeline, given in UTC, shows how quickly the access spread:

  • November 1, 04:33: The attacker made its first call with the key and began mapping what it could reach, reading stored AWS access keys and Azure service principal secrets.
  • Next 24 hours: Each secret led to the next system, from cloud accounts to a data analytics system, then to a source code access token, then to Kubernetes clusters.
  • November 2, early hours: In under 4 hours, the attacker pushed short-lived branches to several repositories to trigger a build workflow that printed encoded build secrets into job logs, bypassing log masking.
  • November 2, 23:31: The last recorded activity.

In total, the attacker touched about 10 cloud accounts in roughly 44 hours.

Kubernetes backdoor and a hunt for crypto wallets

Using access to one Kubernetes cluster, the attacker deployed a container that mounted the underlying server’s file system with read-only access, enough to harvest that server’s credentials. It also tried to create a new administrative cloud identity, which was blocked. Barracuda removed the container on the evening of November 4, Pacific time, and rotated credentials across every system involved.

On servers where it gained shell access, the attacker ran commands searching for environment variables named WALLET, MNEMONIC, SEED and PRIV, along with Ethereum keystore files, extended private keys, seed phrase files and wallet-related Kubernetes secrets. Barracuda said no wallet material existed on those systems. Kjoelen added that AI service credentials within reach may also have been a target.

An engineer spotted unusual activity on their own account on November 4, after the attacker had already finished.

What it means for Barracuda XDR customers

The attacker had access between November 1 and November 2 to the system holding XDR integration credentials. Barracuda said it found no direct evidence those credentials were viewed or accessed, but Kjoelen acknowledged logging in that environment “was not detailed enough to rule it out.”

Barracuda has directly notified customers using an API integration for Barracuda Managed XDR Cloud or Monitored Endpoint whose credentials could be affected. Those customers are advised to revoke and recreate their integration credentials and review activity in connected systems from November 1, 2026, onward. Organizations that were not notified do not need to act. The company also retired 2 internal automation servers whose credentials were exposed, finding no evidence of attacker activity on them.

The advisory lists indicators of compromise, including 4 IP addresses, user agents such as Mozilla/5.0 (compatible; read-only-audit), resources named with the letter “i” followed by digits, such as i31911, and GitHub workflow files named i19ghci.yml and i19ci2.yml.

ARTEX fingerprints, but no attribution

Barracuda said the resource naming, the order of operations and the built-in credential and wallet hunting routines all match ARTEX. Because many groups use the framework, Kjoelen wrote, the match shows “how the attacker operated, not who it was.”

ARTEX has been tied to several campaigns this fall. Microsoft has linked it to the China-state aligned Storm-0940, which Microsoft says was behind the ConnectWise supply-chain attack on Microsoft 365 customers, and to the financially motivated, Cambodia-based Storm-3358. CrowdStrike also reported Vapor Panda using ARTEX AI agents.

On November 13, the FBI released FLASH-20261113-001, warning that ARTEX lets some intrusions progress from initial targeting to privileged access in under 3 hours and citing the ConnectWise Cloud Backup compromise, which it said enabled access to more than 2,250 downstream tenants. The FLASH does not mention Barracuda.

Part of a wider pattern hitting AI companies

Kjoelen compared the incident to the October breach at Railway, where an attacker decrypted customer secrets while hunting for AI API keys, and noted that at least 9 AI developer platforms disclosed incidents in October, a wave that also included Northflank.

Barracuda said it is moving integration credentials to a better-protected store, expanding logging, removing long-lived and over-privileged keys, expiring vendor credentials when engagements end and hardening Kubernetes. The investigation remains ongoing.

Picture of m00s3c

m00s3c

Moose (@m00s3c) is the author of BreachNews, focusing on data breach intelligence, dark web monitoring, and threat analysis. His work involves analyzing breach claims, reviewing leaked datasets, and tracking threat actor activity to provide clear, factual reporting.

Latest News

BREACHNEWS.COM/SUPPORT/

Support Independent News.

Help support breach monitoring, investigations, infrastructure, and reporting.

Support the site →