← Semua berita

Artikel diterbitkan dalam bahasa Inggris.

Threat IntelligenceBackup SecuritySOC

Attackers Hid Webshells in PNG Images on Three Backup Servers and Went Looking for Veeam Credentials. Would You Notice Defender Exclusions on the Server That Holds Your Backups?

Obiguard Research Team·16 November 2026·4 min read

Attackers who exploited two flaws in the AhsayCBS backup management console left five families of webshells, a cryptominer, a Python implant and credential-theft tooling on three servers in one environment, and appear to have been interested in the backup infrastructure itself. The findings come from a Huntress write-up published on 9 November 2026, by Jonathan Semon. The two flaws are CVE-2026-105133, an authentication bypass, and CVE-2026-105134, unauthenticated remote code execution as SYSTEM. This article is based on the Huntress post; we have not independently verified it.

What is known

According to Huntress, the servers were breached between 8 and 11 October 2026, before Ahsay released hotfixes. This is a follow-up to an earlier Huntress report on the same flaws, which covered a webshell and an XMRig miner. The new analysis covers the tooling recovered from three AhsayCBS servers and goes well beyond mining:

  • Five webshell families (JSP). They include nsmind.jsp, a signed file-transfer API, and AhsayCBS.jsp, a loader that hides encrypted payloads in the pixel data of PNG images and runs them from memory.
  • A second miner. A second XMRig miner restarts every minute through a SYSTEM scheduled task, with a wallet and mining pools that differ from those in the first report.
  • A host-bound Python implant. It needs a marker file named after the SHA-256 of the machine's MachineGuid, then runs an embedded x64 DLL in memory that contacts a command-and-control server. Huntress did not observe it running and did not identify the malware family.
  • Supporting tools. The servers also held tunnelling tools, remote-access software, broad Microsoft Defender exclusions, and a command consistent with NetExec's Veeam credential-theft module.

That last item is why Huntress suggests the operator may have been targeting backup infrastructure in particular. Huntress also found no logs proving the attackers used the console's download feature to steal data. Other outlets have repeated the same findings; Huntress is the primary source.

Why this one is different

We have written before about attackers destroying backups before encrypting and about a supply-chain attack on a cloud backup vendor. Here the backup server is not just collateral damage. It is a host that holds credentials for everything else, runs with high privilege, and is often managed less closely than a domain controller or a laptop.

The payload hiding is also worth noting. A webshell stored as image data will not look like a script to a file-type check, and running from memory leaves little on disk. The traces that remain are behavioural: a scheduled task that restarts a process every minute, a registry Run key launching a Python interpreter from a SYSTEM profile, and exclusions added to Defender.

What to check now

  1. Find every AhsayCBS instance. Include ones set up by an MSP, an acquisition or a past project, and confirm each is on a fixed build per Ahsay's own advisory.
  2. Take the console off the open internet. If staff or clients only need it from known networks, restrict it to those.
  3. Hunt on the servers, even patched ones. Look for unexpected .jsp files in the application directories, an embedded Python distribution under the SYSTEM profile, a main.py launched by a registry Run key, and SYSTEM scheduled tasks that run every minute.
  4. Review Microsoft Defender exclusions. Broad, recent exclusions on a backup server are a strong signal on their own.
  5. Treat stored backup credentials as exposed if any instance was reachable and unpatched in early October: rotate the service accounts the backup software uses, including any related Veeam accounts.
  6. Check outbound traffic. Huntress's reports list a command-and-control domain; compare against the indicators in the original post rather than copying them from a summary.

How Obiguard SOC fits

The signals in steps 3 to 6 are small and easy to miss because each one is routine on its own: a scheduled task, a registry key, a Defender setting. Obiguard SOC streams logs and metrics from the systems you already run through OpenTelemetry, so host events from a backup server can be reviewed alongside the rest of your telemetry instead of in a console nobody opens. Alerts arrive with a remediation plan split into now, this week and long-term, so a finding like a new SYSTEM task on a backup host comes with a next step.

Scope matters. SOC works from the telemetry you connect to it. It cannot see events that are not logged, it does not patch AhsayCBS, and it does not remove a webshell or rotate your credentials for you. To see how it would handle logs from your backup servers, get in touch or read about Obiguard SOC.

Sources

Bagaimana Obiguard membantu

Ubah ini menjadi kebijakan yang ditegakkan, bukan sekadar kesadaran.

Obiguard berada di depan setiap permintaan AI organisasi Anda — menyaring prompt dan output berdasarkan guardrail, kerangka kepatuhan, dan jejak audit yang diperlukan oleh kisah seperti ini.

Lihat cara kerjanya →