The Daily Byte
Breaking
Welcome to The Daily Byte — real stories from across Pakistan.
Saturday, 21 November 2026 · Pakistan

The Daily Byte

Stories from Pakistan that deserve to be heard

Barracuda Cloud Breach Traced to Unrevoked AI Key

By The Daily Byte ·16 Nov 2026 ·4 min read
Barracuda Networks headquarters building with the company logo sign outside

Security firm Barracuda has revealed how an attacker used a cloud access key it forgot to revoke to roam across about 10 of its cloud accounts in roughly 44 hours, using an AI-driven attack framework. The intrusion, on 1 and 2 November, was described in a company post-mortem published on 11 November and reported more widely on 15 November.

The key had been issued to a third-party AI service provider for an earlier project and was never retired when the work ended. In October the vendor's own development environment was compromised, and the attacker found the key there; Barracuda later confirmed it had been stored in a configuration file inside the vendor's GitHub repository.

A 44-hour intrusion at machine speed

Barracuda's account, written by its chief information security officer, sets out the attack minute by minute. The first call using the stolen key came at 04:33 UTC on 1 November. Within minutes the attacker was mapping what the key could reach and reading stored secrets, including AWS access keys and Azure service principal secrets.

Over the next 24 hours it used each secret it found to open the next system, moving from Barracuda's cloud accounts to a data analytics system, then to a source-code access token, and then to Kubernetes clusters. In the early hours of 2 November it pushed short-lived branches to a handful of software repositories to trigger build workflows that printed the build system's secrets into job logs, encoded so the usual secret-masking would not hide them. The last activity was recorded at 23:31 UTC on 2 November.

The attacker also left a way back in, deploying a container inside a Kubernetes cluster that mounted the underlying server's file system. Read-only access was enough to harvest that server's credentials. It also tried to create a new administrative cloud identity, an attempt Barracuda says was blocked. The container was removed on the evening of 4 November, and credentials were rotated across every affected system.

What the attacker was looking for

The attacker had two targets. The first was credentials and secrets of every kind: it read secret stores, decrypted infrastructure-as-code state files that often contain credentials, and pulled credentials from server metadata services and Kubernetes service accounts.

The second was cryptocurrency. On servers where it gained shell access, its commands searched explicitly for environment variables named WALLET, MNEMONIC, SEED and PRIV, for Ethereum keystore files, extended private keys and seed phrase files, and for Kubernetes secrets with wallet-related names. Barracuda says no wallet material exists on those systems, but the intent was clear.

Barracuda believes a financially motivated group was behind the intrusion, and notes AI service credentials may also have been part of the prize.

An AI framework did the moving

The probe names, the order in which the attacker worked through Barracuda's systems, and its built-in routines for hunting credentials and wallets all match ARTEX, an open-source AI-driven penetration-testing framework released in July 2026. In ARTEX, a planner AI agent decides the next step while worker agents run common tools and add what they find to a shared map of the target.

Built for authorised testing, ARTEX was reported within weeks of release by researchers at Wiz, Microsoft, Google and CrowdStrike as being used by criminal and state-aligned groups in real attacks. Barracuda stresses the match shows how the attacker operated, not who it was.

Railway, a cloud hosting platform, suffered a similar attack in October, with attackers using stolen AI-service credentials to run up an estimated $740,000 in unauthorised computing costs.

What it means for Barracuda customers

One system the attacker reached stores the credentials Barracuda's extended detection and response platform, known as XDR, uses to collect data from customers' systems. Barracuda found no direct evidence the attacker viewed or accessed those credentials, but logging there was not detailed enough to rule it out.

The company therefore assumed the worst, treated every credential within the attacker's reach as potentially compromised, and contacted the customers concerned directly, asking them to revoke and recreate their XDR integration credentials and review activity from 1 November onwards.

Barracuda learned of the breach on 4 November, when an engineer noticed unusual activity on their own account and the AI vendor contacted the company about its own incident the same day. The intrusion is separate from Barracuda appliance flaws patched around the same time, which the company says do not appear to have been used here.

Barracuda's conclusion is blunt: detect-and-respond as the main strategy is dead when attackers move this fast. Its advice is to remove long-lived standing credentials, especially those held by third parties, shrink the blast radius so no single credential opens the next door, and be able to assume compromise and recover quickly.

Sources: Barracuda Networks blog; Risky Business; Barracuda Trust Center advisory.