Xen Hypervisor EUVD-2026-98824

Exploitation requires: (1) the host CPU to be an AMD Zen2 processor susceptible to Zenbleed (CVE-2023-20593); (2) a Xen version containing the reworked XSA-433 workaround affected by the inverted conditional (fixed in XSA-515), where the DE_CFG chicken-bit mitigation is erroneously left unset; and (3) the attacker to have code-execution capability inside a guest domain on that host. Limiting factors: this is a local, same-host attack (not remote/network), it is confined to vulnerable AMD Zen2 hardware so Intel and other AMD generations are unaffected, and exploitation depends on microarchitectural timing/race conditions that make reliable extraction non-trivial. The inverse condition also means the workaround may be wrongly applied where unneeded, a functional/robustness concern rather than an exposure.

The input CVSS of 9.8 (AV:N/AC:L/PR:N/UI:N/C:H/I:H/A:H) is sharply inconsistent with the vulnerability's actual nature and should be treated with caution. This is a regression in Xen's workaround for Zenbleed (XSA-433 / CVE-2023-20593), an AMD Zen2 microarchitectural data-sampling flaw; an inverted conditional (addressed in XSA-515, CWE-670 incorrect-control-flow) causes the chicken-bit mitigation to be mis-applied, re-exposing the underlying Zenbleed information leak. Zenbleed is a LOCAL cross-domain information-disclosure issue, not a network-exploitable RCE, so the network vector (AV:N) and the Integrity/Availability 'High' impacts in the supplied vector are almost certainly wrong - the real impact is confidentiality only, and the tag 'Information Disclosure' corroborates this. CVSS vector contradicts the description - the description implies local guest execution on AMD Zen2 hardware, so verify the vector with the Xen advisory. EPSS is low at 0.16% (5th percentile), there is no CISA KEV listing and no public exploit identified, which is consistent with a hardware-constrained, non-wormable issue rather than a mass-exploitation priority. Net assessment: a genuine confidentiality risk in multi-tenant AMD Zen2 virtualization environments, but materially lower real-world priority than the 9.8 label suggests.

A malicious or compromised guest VM running on a Xen host backed by an AMD Zen2 processor executes a Zenbleed-style register-rename race to sample stale data from the shared physical register file, because the reworked mitigation was silently disabled by the inverted conditional. Over repeated iterations the guest reconstructs secrets (keys, memory contents) belonging to the hypervisor or co-resident guests. No public exploit is identified for this specific regression, though the original Zenbleed technique (CVE-2023-20593) is well documented.