Threat reportCloud SecurityTL-2026-6219
44-Hour Machine-Speed Cloud Intrusion at Barracuda via Leaked Vendor Access Key, Linked to ARTEX AI Pentest Framework
An attacker used a valid, never-retired cloud access key issued to a third-party AI vendor (whose dev environment was compromised in October) to chain credentials across about 10 cloud accounts, a data analytics system, source-code repositories and Kubernetes clusters between Nov 1 04:33 UTC and Nov 2 23:31 UTC.
Key facts
| Identifiers | TL-2026-6219 |
|---|---|
| Also known as | Barracuda 44-hour cloud intrusion, Anatomy of a machine-speed intrusion |
| Severity | HIGH |
| First disclosed | |
| Affected products | Barracuda Networks Barracuda internal cloud environment (AWS, Azure, analytics system, source-code repositories, Kubernetes); Barracuda Networks Barracuda Managed XDR Cloud / Monitored Endpoint API integrations |
| Weaknesses (CWE) | CWE-798, CWE-613, CWE-532 |
| Tools | ARTEX, Barracuda |
| MITRE ATT&CK techniques | T1027 Obfuscated Files or Information, T1059 Command and Scripting Interpreter, T1059.004 Command and Scripting Interpreter: Unix Shell, T1078.004 Cloud Accounts, T1083 File and Directory Discovery, T1119 Automated Collection, T1136.003 Cloud Account, T1199 Trusted Relationship, T1213 Data from Information Repositories, T1526 Cloud Service Discovery, T1552.001 Credentials In Files, T1552.004 Unsecured Credentials: Private Keys and 6 more in the attack flow |
| Motivation | FINANCIAL |
| Target sectors | technology, cybersecurity, managed-security-services, cloud |
| Target regions | Global |
| Detection rules | 9 (1 shown free) |
| Sources | 8 cited sources |
Detection coverage
As of 2026-11-17, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-6219 across Splunk SPL, Microsoft KQL and Sigma, covering 18 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.
Free detection: Kubernetes Pod Created with hostPath Host Filesystem Mount (Barracuda Backdoor Container Pattern)
TL-2026-6219: Detects creation of a Kubernetes pod mounting a host filesystem path (hostPath), including read-only mounts. The Barracuda attacker deployed a read-only host-mounting container to harvest the node's credentials and Barracuda treats it as the persistence mechanism (T1610, T1552.007). Read-only does not make this benign: credentials on the host remain readable.
- Languages: Splunk SPL, Microsoft KQL, Sigma
- ATT&CK: T1610, T1552.007
- Data sources: kubernetes_audit (index=kubernetes sourcetype=kube:apiserver-audit), AKS kube-audit (AzureDiagnostics)
- Severity: high
- Confidence: high
Splunk SPL
index=kubernetes sourcetype="kube:apiserver-audit" verb=create objectRef.resource=pods requestObject.spec.volumes{}.hostPath.path=*
| search NOT objectRef.namespace IN ("kube-system","kube-node-lease")
| eval host_paths=mvjoin('requestObject.spec.volumes{}.hostPath.path',","), images=mvjoin('requestObject.spec.containers{}.image',",")
| eval root_mount=if(match(host_paths,"(^|,)/(,|$)"),"yes","no")
| table _time user.username sourceIPs{} objectRef.namespace objectRef.name host_paths images root_mount userAgentMicrosoft KQL
AzureDiagnostics
| where TimeGenerated > ago(24h)
| where Category == "kube-audit"
| extend a = parse_json(log_s)
| where tostring(a.verb) == "create" and tostring(a.objectRef.resource) == "pods"
| extend Volumes = tostring(a.requestObject.spec.volumes)
| where Volumes has "hostPath"
| where tostring(a.objectRef.namespace) !in ("kube-system", "kube-node-lease")
| project TimeGenerated, User = tostring(a.user.username), Namespace = tostring(a.objectRef.namespace), Pod = tostring(a.objectRef.name), SourceIps = tostring(a.sourceIPs), Volumes, Images = tostring(a.requestObject.spec.containers), UserAgent = tostring(a.userAgent)Sigma
title: Kubernetes Pod Created with hostPath Volume Mount
id: 2d9b7c40-8a13-4f6e-b5c2-71e4a9d03f22
status: experimental
description: Detects Kubernetes API audit events where a pod is created with a hostPath volume, the persistence and host credential harvesting technique used in the Barracuda 44-hour intrusion (TL-2026-6219), where a read-only host filesystem container was deployed.
references:
- https://blog.barracuda.com/2026/11/11/anatomy-of-a-machine-speed-intrusion
- https://gbhackers.com/44-hour-cloud-attack/
author: Threadlinqs Team
date: 2026-11-17
tags:
- attack.execution
- attack.t1610
- attack.credential-access
- attack.t1552.007
logsource:
category: application
product: kubernetes
service: audit
detection:
selection:
verb: 'create'
objectRef.resource: 'pods'
selection_hostpath:
requestObject.spec.volumes|contains: 'hostPath'
filter_system:
objectRef.namespace:
- 'kube-system'
- 'kube-node-lease'
condition: selection and selection_hostpath and not filter_system
falsepositives:
- Node agents and DaemonSets that require host access
- Cluster administrators debugging nodes
level: high
Known false positives:
- Node-level agents (log shippers, CSI drivers, monitoring DaemonSets, EDR) legitimately mount hostPath
- Platform engineers running debug pods on nodes
Tuning notes: Exclude approved DaemonSet namespaces/service accounts via a lookup; raise severity for hostPath of /, /var/lib/kubelet, /etc, /root or /var/run/docker.sock, or when the requestor is a non-human identity; enforce Pod Security 'restricted' to block rather than only alert.
- Azure Key Vault Mass Secret Read by Service Principal (Credential Chaining, Barracuda Intrusion) · unlock with TL Creds
- Cloud Administrative Identity Creation Attempt or ARTEX-Style i-Digits Naming (AWS IAM) · unlock with TL Creds
- Cloud Instance Metadata Credential Retrieval from Shell or Script (IMDS) · unlock with TL Creds
- GitHub Workflow [indicator] / [indicator] CI Secret Exfiltration to Job Logs · unlock with TL Creds
- Linux Shell Hunting for WALLET MNEMONIC SEED PRIV Variables and Crypto Wallet Files · unlock with TL Creds
- Terraform State File Read Followed by KMS Decrypt by Same Access Key (IaC State Credential Theft) · unlock with TL Creds
- ARTEX-Style Long-Lived Cloud Key Discovery and Secret Harvest Burst (AWS) - Barracuda 44-Hour Intrusion · unlock with TL Creds
- Kubernetes Secrets Sweep and Wallet-Named Secret Access by One Identity (Credential Chaining) · unlock with TL Creds
Indicators of compromise
As of 2026-11-17, Threadlinqs Intelligence tracks 18 indicators of compromise for TL-2026-6219.
Sample indicators (3 of 18)
- IP address:
180.178.56[.]10 - IP address:
52.128.238[.]126 - IP address:
221.146.68[.]107
Defanged, TLP:CLEAR. Full set of 18 indicators: unlock with TL Creds.
Technical analysis
Barracuda linked the methods to ARTEX, an open-source AI-driven penetration-testing framework, based on resource naming, operational sequencing and automated credential-hunting routines; no specific threat group is named.
Barracuda disclosed (Trust Center advisory Nov 10, 2026; technical blog Nov 11, 2026) an intrusion into its cloud environment that lasted roughly 44 hours. No software vulnerability was exploited and no CVE is involved. Initial access was a valid cloud access key that Barracuda had issued to a third-party AI vendor for an earlier project and never retired. The vendor's development environment was compromised in October 2026; GBHackers reports the key was taken from a configuration file in the vendor's GitHub repository, and Barracuda states this is where the attacker most likely obtained it.
The first recorded request with the key was at 2026-11-01 04:33 UTC. Within minutes the attacker mapped what the key could reach and read stored secrets, including AWS access keys and Azure service principal secrets. Over about the next 24 hours each harvested secret was used to reach the next system: roughly 10 cloud accounts, a data analytics system, a source-code access token, and Kubernetes clusters. The attacker decrypted infrastructure-as-code state files, pulled credentials from server metadata services and Kubernetes service accounts, and read secret stores. Resources created by the attacker used machine-style labels consisting of the letter 'i' followed by digits (e.g. i31911, i28324probe).
In the early hours of Nov 2, over just under four hours, the attacker pushed short-lived branches to several repositories to trigger a build workflow (workflow files i19ghci.yml and i19ci2.yml are cited) that printed build-system secrets into the job log in an encoded form so that the CI secret-masking would not redact them. In Kubernetes the attacker deployed a container that mounted the host server's file system; it was read-only, but sufficient to harvest the host's credentials, and Barracuda treats it as the persistence mechanism. The attacker also tried to create a new administrative cloud identity; that attempt was blocked. On servers where it obtained shell access, the attacker's commands searched for environment variables named WALLET, MNEMONIC, SEED and PRIV, Ethereum keystore files, extended private keys, seed-phrase files and Kubernetes secrets with wallet-related names. No wallet material existed on those systems. Barracuda also notes AI-service credentials may have been a target and does not rule out that motive. Last recorded attacker activity was Nov 2 at 23:31 UTC.
An engineer noticed unusual activity on their own account and reported it on Nov 4; the same day the AI vendor contacted Barracuda about its own incident. Barracuda removed the backdoor container that evening (Pacific) and began credential rotation across the environment. One reached system stores credentials used by Barracuda XDR for data collection; Barracuda found no direct evidence they were accessed but its logging could not rule it out. Customers with potentially affected Managed XDR Cloud or Monitored Endpoint API integration credentials were contacted directly and asked to revoke and recreate them and to review connected-system activity from Nov 1. Barracuda says organizations it did not contact need take no action.
Attribution: Barracuda linked the activity to ARTEX based on probe/resource naming, the order of the attacker's work, and built-in automated credential- and wallet-hunting routines. ARTEX is an open-source, LLM-driven, multi-agent penetration-testing framework (a planner agent directs worker agents that run common tools and update a shared target map), reported as released in July 2026 and widely forked. Barracuda stresses that many groups use ARTEX, so the match identifies a tool/operating method, not a specific actor, and does not prove fully autonomous execution. Railway's separate October 2026 incident (stolen AI-provider credentials, ~$740,000 unauthorized usage) is cited only as a parallel, not as evidence of shared attribution. No IP addresses, domains or file hashes were published, so no BeaconBeagle network correlation was possible.
MITRE ATT&CK techniques by tactic:
Defense Evasion
T1027 Obfuscated Files or Information
Execution
T1059 Command and Scripting Interpreter; T1059.004 Command and Scripting Interpreter: Unix Shell; T1610 Deploy Container
Initial Access
T1078.004 Cloud Accounts; T1199 Trusted Relationship
Discovery
T1083 File and Directory Discovery; T1526 Cloud Service Discovery; T1580 Cloud Infrastructure Discovery
Collection
T1119 Automated Collection; T1213 Data from Information Repositories
Persistence
Credential Access
T1552.001 Credentials In Files; T1552.004 Unsecured Credentials: Private Keys; T1552.005 Cloud Instance Metadata API; T1552.007 Container API; T1555 Credentials from Password Stores; T1555.006 Credentials from Password Stores: Cloud Secrets Management Stores
Remediation
Immediate actions
- Inventory and revoke access keys, tokens and service principals issued to third parties or to completed projects
- Rotate all secrets reachable from any exposed credential: AWS keys, Azure service principal secrets, source-code tokens, Kubernetes secrets, CI/CD secrets and AI-service keys
- Barracuda XDR customers contacted by Barracuda: revoke and recreate Managed XDR Cloud / Monitored Endpoint API integration credentials and review connected-system activity from 2026-11-01
- Hunt for repository branches that were pushed and deleted quickly alongside CI runs whose logs contain encoded secret output
- Audit Kubernetes for pods mounting hostPath volumes (including read-only) and for unexpected workloads
Workarounds
- Enforce Kubernetes admission controls (e.g. Pod Security restricted profile) blocking hostPath mounts
- Require IMDSv2 and restrict metadata service access from workloads
- Alert on creation of new administrative cloud identities
Longer-term hardening
- Replace long-lived credentials with short-lived, narrowly scoped credentials
- Limit blast radius so no single credential unlocks the next system
- Store integration secrets in storage protected against cluster-administrator compromise
- Expand audit logging so access to sensitive systems can be confirmed or ruled out
- Build the ability to rotate secrets at scale and rebuild from known-good state
- Restrict who can create workflow runs from unreviewed branches and do not rely on log secret masking
Affected and fixed versions
- Barracuda Networks — Barracuda internal cloud environment (AWS, Azure, analytics system, source-code repositories, Kubernetes)
Vulnerable versions: Environment accessible via the unretired third-party vendor access key
Fixed in: Credentials rotated and backdoor container removed 2026-11-04 - Barracuda Networks — Barracuda Managed XDR Cloud / Monitored Endpoint API integrations
Vulnerable versions: Integration credentials stored in a reached system; access not ruled out
Fixed in: Customers notified directly to revoke and recreate credentials
Timeline
Last updated · revalidated 1× · latest source
- ARTEX open-source AI-driven penetration-testing framework published on GitHub (Help Net Security gives July 26; sources differ slightly on exact timing)
- Over roughly the next 24 hours harvested secrets are chained into ~10 cloud accounts, a data analytics system, a source-code access token and Kubernetes clusters; IaC state files decrypted and metadata/service-account credentials pulled
- 04:33 UTC: first recorded request using the valid third-party AI vendor cloud access key; attacker maps reachable resources and reads stored secrets including AWS keys and Azure service principal secrets
- 23:31 UTC: last recorded attacker activity; a Kubernetes container mounting the host filesystem had been deployed and an attempt to create an administrative cloud identity was blocked
- Early hours: over just under four hours, short-lived branches pushed to several repositories to trigger a build workflow that prints build-system secrets to job logs in encoded form to bypass masking
- An engineer reports unusual activity on their account and the AI vendor informs Barracuda of its own compromise; Barracuda removes the backdoor container and begins credential rotation
- Barracuda publishes Trust Center advisory and discloses the incident; affected Managed XDR Cloud / Monitored Endpoint customers contacted to recreate integration credentials
- Barracuda publishes 'Forty-four hours: Anatomy of a machine-speed intrusion' linking the activity to ARTEX tooling
- Barracuda updates the Trust Center advisory with root cause, published indicators (IPs, user agents, resource and workflow names), retirement of two internal automation servers with exposed credentials (no attacker activity found) and the ARTEX resemblance; same day the FBI issues a FLASH alert on ARTEX agents automating network attacks.
- GBHackers publishes coverage including resource names and workflow files associated with the intrusion
Update history
- 2026-11-17 — Exposed AI Vendor Cloud Key Used to Breach ~10 Barracuda Cloud Accounts in 44 Hours (ARTEX-linked): What changed No field escalation; severity HIGH and exploitability ACTIVE already match. Record is enriched with details from the Nov 13 advisory update. New indicators (13) Indicator values are withheld at this access level.
Sources cited
- Forty-four hours: Anatomy of a machine-speed intrusion (Barracuda) blog.barracuda.com · Vendor report · 2026-11-11
- Check Point Details Autonomous ARTEX Attacks by Cybercriminals letsdatascience.com · Research
- ARTEX AI Framework Linked to 44-Hour Barracuda Cloud Attack Targeting Credentials and Crypto Wallets (GBHackers) gbhackers.com · News · 2026-11-17
- Help Net Security: ARTEX AI Pentest Framework Hits 12K GitHub Forks, Already Used Against 200+ Orgs in 11 Days aiweekly.co · News
- Dark Reading: ARTEX AI Pen Test Tool Hackers Weapon Dark Reading · News
- Open-Source ARTEX Turns AI Agents Into Autonomous Hacking Teams (AlphaSignal) alphasignal.ai · News
- Railway Breach Leads to $740,000 in Fraudulent AI Charges (BleepingComputer) - parallel incident, not shared attribution BleepingComputer · News
- Hackers stole AI credentials to get $740,000 in free compute, Railway picked up the bill (CSO Online) csoonline.com · News
Drafted by the Threadlinqs research pipeline from 8 sources; published ; last updated .