Mior and Skyfall Attacks Expose AMD SEV-SNP VMs Allowing Rogue Hosts to Steal Secrets and Modify Memory

Two new attacks named Mior and Skyfall were disclosed this week that enable a malicious hypervisor to bypass AMD SEV-SNP protections on virtual machines. Mior, developed by researchers at ETH Zurich, corrupts the Reverse Map Table during VM startup by redirecting PSP writes through a controlled PCIe device, ultimately allowing memory reads, writes, and forged attestation reports. Skyfall, a joint effort by ETH Zurich and Google, leverages Transient Scheduler Attacks to extract register values and disk encryption keys from failed VM launches without requiring BIOS modifications. Both attacks were demonstrated on AMD EPYC processors across Zen 3, Zen 4, and Zen 5 generations, with AMD issuing corresponding CVEs and firmware updates scheduled through December 2026. The flaws highlight persistent challenges in confidential computing even when hardware memory encryption is enabled.

The Hacker NewsTranslated from English

Read at source

Two attacks made public this week allow a malicious host to read data inside AMD SEV-SNP-protected virtual machines (VMs), which are built to keep the server's operator out. One of them, Mior, can also modify a VM's memory and fake the signed report confirming which software it runs.

Both need control of the server's hypervisor, and Mior also needs control of the server's BIOS. Researchers showed both in lab tests on AMD EPYC server processors. AMD lists SEV firmware for Mior and microcode for Skyfall, both dated August 12. It plans to give server makers platform firmware in November for Skyfall and in December for Mior.

Mior Attack Details

The ETH Zurich team behind Mior attacked SEV-SNP while it was starting up. During start-up, the Platform Security Processor (PSP) fills in the Reverse Map Table (RMP). The team redirected the PSP's writes to the memory of a graphics card it controlled. Start-up then finished without errors, but part of the real RMP stayed uninitialized, allowing the hypervisor to rewrite protections.

The team then enabled debugging on a confidential VM and used the PSP's own debug commands to read and write the VM's memory. This worked in all 1,000 attempts. In a second test, the team copied a harmless VM's measurement into a malicious VM so attestation reports described the harmless image instead.

Skyfall Attack Details

Skyfall comes from a team at ETH Zurich and Google. It builds on Transient Scheduler Attacks (TSA). The team found that when a hypervisor intentionally causes a VM to fail to start, the processor has already cached part of the VM's saved state. This way, the team recovered a VM's saved register values and a disk-encryption key from a test VM.

Skyfall can also read VM memory, but only three fixed cache lines in a page it targets. The attack needs a malicious host or hypervisor but no tampering with the hardware. The team tested it on EPYC 7313 (Zen 3) and EPYC 9124 and 9334 (Zen 4) chips.

Affected Processors and Fixes

AMD tracks Mior as CVE-2026-0462 and Skyfall's VM leak as CVE-2026-40671. The company lists the same five EPYC server lines for both flaws, with platform firmware updates targeted for November and December 2026. The EDK II project assigned Skyfall's SMM leak CVE-2026-18448.

Because the fixes are implemented in the host's firmware, customers who rent confidential VMs depend on their cloud provider to install them. Both teams have published proof-of-concept code.

Meridian digestThe week's most consequential security stories, distilled into one briefing.See the digest

More in Vulns & Exploits

Check Point Security Gateways Compromised via PHP2Shell Exploits to Deploy VMware ESXi Data Wiper
Microsoft November 2026 Security Updates Break Printing for Complex Documents
Check Point Releases Emergency Hotfixes for Zero-Day CVE-2026-123039 Exploited for Root Access