A team of researchers, which included representatives from four US universities, studied the technology of Secure Encrypted Virtualization (SEV) from AMD and found that under certain circumstances, attackers can bypass its protection.
The research findings were presented this week at the ACM Asia Conference on Computer and Communications Security in Auckland, New Zealand.In a report titled “The SEVerESt of Them All: Inference Attacks Against Secure Virtual Enclaves” scientists described two attack methods.
The technicians presented by them allow unscrupulous administrators of cloud servers or attackers who hacked the hypervisor to detect applications running on the virtual machine that are protected by SEV, as well as to inject and extract data from virtual mach3.bn/ines.
Read also: Microsoft closed 77 vulnerabilities as part of Patch Tuesday
As AMD explains, SEV protects guest virtual machines both from each other, and from the software running on the host and its administrators. No matter what happens on the same virtual machine, this should not affect any other virtual machines, the host operating system, the hypervisor, or the administrators.
However, researchers have demonstrated that SEV technology is not capable of repelling attacks from a malicious hypervisor.
“Passively watching the changes in the logs, an attacker can recover critical information about actions in encrypted guest systems”, – says researchers’ report.
The attack even works with Secure Encrypted Virtualization Encrypted State (SEV-ES), an advanced memory protection technique that encrypts not only the RAM, but also the virtual machine control unit.
This block is a memory area where the contents of virtual machine’ CPU registy are saved when it is forced to yield to the hypervisor. In theory, encryption should prevent the hypervisor from understanding the context of the suspended virtual machine, but researchers have proven the opposite.
You can get acquainted with the details of the attacks here.