Barracuda Hacked as Artex-Powered Cyberattacks Multiply

Barracuda Hacked as Artex-Powered Cyberattacks Multiply
3 min read

The Artex framework, designed for AI-assisted penetration testing, has become a common tool in the arsenal of Chinese malicious actors, and Barracuda is among its victims. The FBI has issued a new alert dedicated to this tool.

Published on GitHub in July 2026, Artex relies on a planner, responsible for formulating an objective and creating tasks, and on agents that execute these tasks. The platform itself records progress, results, and assets. It originally used DeepSeek v4.1-flash, but can also call upon other LLMs (large language models). GLM-5.3 and Grok 4.6 have been observed in some instances. Its users, however, rely mostly on Chinese models, which are more permissive than American models.

Designed for security testing, the tool has been used in real intrusions since August at least, and its use in attacks has increased sharply in recent weeks. In its alert, the FBI cites a key passage: ‘The framework allows malicious actors with minimal technical expertise to simultaneously conduct complex operations targeting multiple victims, fundamentally accelerating the speed of attack and expanding operational scale.’ The agency mentions its use in attacks against South Korean financial institutions and against ConnectWise. It also discusses the risks of faster attacks through automation, the exploitation of neglected peripheral equipment, and the search for attack vectors in the supply chain.

The developer removed its GitHub repository in October, but Artex continues to spread through other channels. In early November, Censys identified about 1,500 hosts exposing Artex consoles. In October, the tool was used by the Chinese espionage group Vapor Panda in a campaign targeting over 200 organizations in 14 countries, including ‘ministries of foreign affairs, a European presidency, and American pharmaceutical, semiconductor, defense, and nuclear organizations,’ according to SentinelOne. Storm-0940, another Chinese state-sponsored group, also uses Artex. Evidence also shows that profit-motivated actors, like Storm-3358, have adopted it, according to Talos, Cisco’s intelligence division.

An Intrusion at Barracuda

Once access was obtained, attackers were observed abusing compromised credentials and keys, elevating their privileges, installing persistence mechanisms, moving laterally, and stealing credentials and data. Talos describes impacts ranging ‘from access to cloud secrets and email data to using compromised applications to reach connected environments.’ Amazon makes similar observations and notes that one of the most notable Artex campaigns aimed at large-scale theft of credentials providing access to AI models, while another targeted cryptocurrency exchange platforms.

Barracuda, victim of an Artex-powered intrusion, published similar findings in a blog post. The attackers used a valid and exposed cloud access key to map its perimeter, then reused discovered secrets to access other systems. The company writes: ‘In total, the attacker touched about 10 cloud accounts in roughly 44 hours. The work was methodical and tireless.’ The attack took place between November 1st and 2nd. It targeted secrets, configurations, Kubernetes service accounts, and server metadata services, and left a backdoor in a Kubernetes cluster. On some servers, the attackers obtained shell access and searched for Ethereum keystore files, recovery phrase files, and other information related to cryptocurrency wallets. Barracuda links this attack to the October intrusion against the Railway hosting platform, as well as the wave of intrusions reported at dozens of AI development platforms. The company did not attribute the attack to a specific group, noting that Artex is used by many actors.

In its analysis, Amazon emphasizes that attackers relied on long-known techniques, but at a much higher speed thanks to AI. The group states: ‘The agents succeeded where basic security hygiene was lacking, including default or hard-coded credentials, keys left in code and public artifacts, unauthenticated inference interfaces and access points, and unpatched software.’ Daniel Shechter, co-founder and CEO of Miggo Security, believes that ‘autonomous attackers don’t need a zero-day.’

S’abonner
Notification pour
guest

0 Commentaires
Le plus ancien
Le plus récent Le plus populaire
0
Nous aimerions avoir votre avis, veuillez laisser un commentaire.x