
Indicators and a read-only hunt guide for VMware vCenter CVE-2026-123890, from Google's report. One source, not tested, will go stale. Check the dates.
Read first. This package is aggregated from public research and vendor documents by SOC-Lite by BansalAI. It is not tested by us, it comes with no warranty, and it is not professional advice. It is as of 2026-11-19, the day the flaw was published, and it will go stale. See DISCLAIMER.md. We are not affiliated with any company credited in CREDITS.md, and none of them reviewed or endorsed this package.
On 2026-11-19 Broadcom published advisory VMSA-2026-0009. One flaw in it, CVE-2026-123890, lets someone with network access to VMware vCenter get in without a password. Broadcom lists no workaround. The same day, Google's threat researchers (Google Threat Intelligence Group and Mandiant) said a suspected China-linked group had used the flaw since early September 2026, before any fix existed. They describe one break-in: the group got an administrator session on vCenter, copied the directory database and the key that signs single sign-on tokens, then moved on to ESXi hosts and copied files out of virtual machines.
This repository copies the indicators the report publishes (with credit) and adds a guide to checking your own systems.
Not technical? Start with the plain-language version: VMware vCenter Update: What to Ask Whoever Runs Your Servers on the SOC-Lite blog. It gives five questions to put to your IT person or provider.
| File | What |
|---|---|
iocs/iocs.csv | One row per indicator, with the source, the date, how fragile it is, and a review date |
docs/hunt-guide.md | How to check your version, the network, vCenter, ESXi hosts and the Linux machines near vCenter, what the results do and do not mean, and what to do if something looks real |
CREDITS.md | The research and documents this is built from, and what we took from each |
DISCLAIMER.md | The full disclaimer |
Rows: 32 · verified (two or more sources): 0 · single source: 32
These counts match iocs/iocs.csv; we check them every time we change the file (a test checks them in our
private working copy). Every value was read from the publisher's own copy of the report. One source means no row is verified
yet. What we took, and where it came from, is in CREDITS.md.
iocs.csv (indicator values, with credit)| Source | Rows | Rows by type | What we took | Link |
|---|---|---|---|---|
| Google Threat Intelligence Group and Mandiant | 32 | 11 ip, 5 filename, 6 file-path, 3 string, 1 account, 6 behaviour | The report's two indicator tables (eleven addresses and six file names), and the paths, ports, account-name prefix, SSH banner, tool file names and described behaviour that the report's text gives | https://cloud.google.com/blog/topics/threat-intelligence/armory-castle-exploits-vsphere |
| Total | 32 |
sps-. The report redacts the
rest, and so do we.iocs.csv| Column | Meaning |
|---|---|
id | Stable row ID (VC-nnnn) |
wave | Always vcenter in this release |
type | ip, filename, file-path, string, account or behaviour |
value | The indicator. Addresses are written as normal text, not defanged. Never open them in a browser |
match | exact, substring, prefix or wildcard |
evidence | Where it would show: network, file, process, access-log or event-log |
fragile | yes if the attacker can change it trivially |
verified | yes only when two or more independent primary sources report it. no means a single source |
source_url, source_date | Where we took it from and the date of that source |
fetched, review_by | When we read it, and when it should be re-checked (30 days for every row here) |
notes | Context and caveats. Sentences starting "Our note" are ours, not the source's |
Search your own systems with the steps in the hunt guide. This package does not ship anything that runs on your systems or collects data from them.
The text we wrote is licensed under CC BY 4.0 (LICENSE). The indicator values and facts belong to the research and documents that reported them: they are credited in CREDITS.md, and we do not relicense them. Details and how to credit us are in NOTICE.md.
Broadcom's advisory is the authority on which versions are affected and fixed. Google's report is the authority on what the group did. Both may be updated: follow the links in CREDITS.md.