Barracuda says hackers using the Artex AI tool roamed 10 of its cloud accounts in 44 hours

Science & Technology

Barracuda: AI Hacking Tool Artex Roamed 10 Cloud Accounts in 44 Hours

Barracuda says Artex users with a leaked vendor key reached 10 cloud accounts in 44 hours, touching the store of XDR customer credentials and prompting rotation calls.

By
Tom Whitfield
Filed
Channel
Science & Technology
Read
4 min read

Barracuda Networks says an attacker wielding Artex, the open-source AI hacking framework the FBI warned about last week, moved through roughly 10 of its cloud accounts in about 44 hours before anyone noticed. The intrusion, which ran from Nov. 1 to Nov. 2, reached a system holding the credentials Barracuda's Managed XDR service uses to connect to customers' Microsoft 365, Google Workspace and other environments, the company disclosed in a Nov. 13 advisory.

The entry point was not exotic. Barracuda had issued a cloud access key to a third-party vendor it describes as an AI company for an earlier project, and never retired the key when the work ended. "That is on us," chief information security officer Arve Kjoelen wrote in the company's technical account. The key sat in a configuration file in the vendor's GitHub repository, and it leaked when the vendor's development environment was compromised in October.

How did the intrusion unfold?

The attacker first used the key at 04:33 UTC on Nov. 1, according to Barracuda's timeline. From there, the intruder worked in steps, using each secret it found to reach the next system: cloud accounts first, then a data analytics system, then a source-code access token, then Kubernetes clusters.

In the early hours of Nov. 2, the attacker pushed short-lived branches to several repositories to trigger a build workflow. That workflow printed the build system's secrets into job logs, encoded so the usual masking would not hide them. The last recorded activity came at 23:31 UTC that day — a window of roughly 44 hours in total.

The intruder also left a backdoor container in a Kubernetes cluster, mounting a server's file system to harvest credentials, and attempted to create a new administrative cloud identity. That attempt failed. Barracuda removed the container on the evening of Nov. 4, Pacific time, after an engineer reported unusual activity on their own account and the vendor separately disclosed its own breach the same day.

The practical risk sits in the credential store. Barracuda said logging there was not detailed enough to rule out access, so it treated every credential within the attacker's reach as compromised and rotated secrets across its environment. A spokesperson told TechCrunch on Monday that the company "directly notified and advised a minority of XDR customers to revoke and recreate integration credentials," without saying how many.

Why does Barracuda blame Artex?

The fingerprints matter as much as the timeline. The intruder named cloud and Kubernetes resources with the letter "i" followed by digits, such as i31911 and i28324probe, and ran built-in routines that searched servers for Ethereum keystore files, seed phrases and wallet environment variables. Barracuda said those naming patterns and the order of operations match Artex, but it has not attributed the attack, because many groups use the tool. TechCrunch described the attackers as financially motivated criminals.

Artex itself is a recent artifact. It was published on GitHub in July and uses a planner agent to set tasks for worker agents. The developer pulled the repository in October, but copies still circulate, and SecurityWeek reported that Chinese state-linked groups including Vapor Panda and Storm-0940 have adopted it alongside criminal crews. The FBI's flash alert on Artex said intrusions using the tool can go from targeting to compromise in under three hours.

Kjoelen drew the broader lesson plainly: "Defenders who rely on spotting malicious activity and then responding to it will often arrive after the attacker has finished." The attacker here finished before the defense started — detection came three days after the first key use.

What should customers do?

Barracuda's advisory applies only to organizations using an API integration for Barracuda Managed XDR Cloud or Monitored Endpoint. Customers that received no notice need take no action. Notified customers should review activity in connected systems dating from Nov. 1.

The company has also retired two internal automation servers whose credentials were exposed, and it says the investigation continues. Barracuda confirmed the breach is separate from the two Email Security Gateway flaws it disclosed as exploited on Nov. 10.

The takeaway for the supply chain is blunt: an unretired key at an AI vendor bought an attacker 44 hours of quiet movement inside a security vendor's own cloud, and no one caught it until an engineer noticed something odd on a personal account.

Source: mlq.ai

Share this article:

More from Tom Whitfield

Tom Whitfield

Show full bio

Staff writer covering companies, deals and strategy in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

999 articles

Related articles

« Previous article