VMware's 9.8 vCenter bypass, and the KVM writeup Digital Ocean pulled
Two critical VMware advisories expose vCenter and SDDC Manager, while a 5.5-rated KVM bug handed one researcher 127GB of host RAM and his writeup vanished within hours.
VMware published two critical advisories covering a 9.8-rated vCenter authentication bypass in Cloud Foundation and a root-level flaw in SDDC Manager, while a separate 5.5-rated KVM bug in AMD nested virtualization let researcher Oren Yomtov reach 127GB of host RAM on a Digital Ocean Droplet. The consequence for decision-makers is that management appliances, not workloads, are now the fastest route to a full environment compromise, and disclosure timing is being negotiated with cloud providers rather than set by researchers.
VMware just handed virtual infrastructure teams two critical-rated advisories, and the worst of them scores 9.8. VMSA-2026-0009 covers CVE-2026-123890, a flaw VMware says means a malicious actor with network access to vCenter may exploit it to bypass authentication and gain unauthorized access to the system. vCenter is the console admins use to manage fleets of virtual machines, and the flaw is present in VMware's flagship Cloud Foundation suite. The same advisory covers CVE-2026-123891, which allows attackers to read and write the database behind SDDC Manager, the tool for managing entire software-defined datacenters, then execute arbitrary commands and elevate privileges to root. Both, at least, are fixed. The virty giant has not managed the same for everything on its list.
Then there is the KVM mess, where the fix landed before the CVE did. CVE-2026-99312, disclosed November 18, lets an attacker use a guest virtual machine to inflict denial of service or memory corruption attacks on the host. That is normally nightmare fuel: access to a single tenant should never let someone harm other guest VMs or peer into the memory a guest uses. This one is narrower. It only impacts users of AMD processors who employ nested virtualization, the practice of running a hypervisor inside a virtual machine, which is uncommon in the enterprise but used by clouds to do things like powering OS emulators. That is why the CVSSv3 score is a modest 5.5. Security researcher Oren Yomtov of Accomplish.ai says the flaw meant he could access 127 GB of host RAM on a Digital Ocean "Droplet," the junior cloud's name for its cloudy VMs.
Yomtov says he told Digital Ocean and the Linux kernel security team about the flaw in October. Maintainers got busy: a fix reached the mainline kernel on November 2, and full fixes landed on November 17. The CVE appeared on November 18. Yomtov and Terra Security quickly posted their analysis of the flaw, then took the posts down within hours. The Register understands that Digital Ocean asked for the posts to come down because it needed more time to update its Linux implementations, and the company has not responded to multiple requests for comment. Read that sequence again, because it is the part with teeth. The technical fix was public more than two weeks before the CVE number existed, and the writeups still disappeared on request from a provider that had not finished patching.
VMware is not done. VMSA-2026-0010 warns that Cloud Foundation, Aria Operations, and VCF Operations HCX all use Apache, which carries CVE-2026-119618, a 9.8-rated remote code execution vulnerability. There is no patched product yet. VMware plans to ship new versions of its products with a patched version of Apache, but for now customers get only workarounds that disable the http2 protocol. That is a genuine trade-off rather than a formality: http2 is load-bearing for performance in modern web stacks, and turning it off to stay safe is the kind of call that escalates to a change board. All of it arrives while admins are already sprinting through emergency patches for PHP, Java, and PostgreSQL.
Context on why the 9.8 rating is not a rounding error. vCenter is the control plane for virtualization estates: it is where credentials, inventory, and management of the hypervisor fleet live. A network-reachable authentication bypass means an attacker does not need a stolen password or a phished admin, just a route to the service. SDDC Manager sits a layer up, orchestrating the software-defined datacenter itself, and root on that box is not a data problem, it is a rebuild-the-environment problem. For any organization running Cloud Foundation, the question this week is not whether the flaws are serious. It is how fast the patch cycle actually completes, and who owns the clock.
Terra's analysis adds the uncomfortable part. The company asserts the KVM flaw was introduced to KVM in 2022, when developers wrote a fix for CVE-2022-3344. Terra believes 3.2 percent of CVEs in the Linux kernel were introduced by the fix for another CVE. That is the compounding cost of open-source security: every remediation is new code, new code is new attack surface, and the dependency graph underneath a modern cloud is deep enough that nobody can hold it in their head. It also explains why disclosure timing gets fought over. Publishing analysis before downstream providers have patched hands attackers a map, so providers push back, and the pushback lands on the researchers rather than on the customers paying the bill.
The mitigation for CVE-2026-99312 is refreshingly cheap: disable nested virtualization in KVM and the flaw is defeated without an update. The catch is that the organizations most exposed are precisely the ones that use nested virtualization on purpose, which is why Digital Ocean's patching clock mattered enough to ask researchers to hold their writeups. For executives, the through-line across all three advisories is asset visibility. You cannot patch a hypervisor you do not know you run, a nested virtualization dependency nobody documented, or an Apache build buried inside a management appliance. This week the security budget question is not whether to spend. It is whether your inventory is good enough to know what you are spending it on, and whether your board would learn about a 127 GB exposure from your team or from a headline.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology
Anthropic mathematician claims Claude helped crack the Riemann hypothesis
If the proof holds, AI has toppled math's most famous open problem; if it does not, it is a case study in announcing breakthroughs before anyone verifies them.
AI's $4 Trillion Wipeout Erases Nasdaq's Gains for the Year
A brutal week took roughly $4 trillion off the AI sector and left the Nasdaq flat for the year, and the damage ran the full length of the supply chain.
Anthropic mathematician says AI proved the Riemann hypothesis. Verification will take time.
A single unverified claim from inside Anthropic now sits against more than 160 years of open mathematics, and the gap between asserting a proof and surviving peer review is the entire story.



