AMD has launched mitigation and firmware updates to deal with a high-severity vulnerability that may be exploited to load malicious CPU microcode on unpatched gadgets.
The security flaw (CVE-2024-56161) is attributable to an improper signature verification weak point in AMD’s CPU ROM microcode patch loader.
Attackers with native administrator privileges can exploit this weak point, ensuing within the lack of confidentiality and integrity of a confidential visitor working beneath AMD Safe Encrypted Virtualization-Safe Nested Paging (SEV-SNP).
In keeping with AMD’s improvement sources, SEV isolates company and the hypervisor from each other, and SEV-SNP provides reminiscence integrity safety that creates an remoted execution atmosphere by serving to stop malicious hypervisor-based assaults (e.g., information replay, reminiscence re-mapping, and extra).
AMD now gives mitigation requiring a microcode replace on all affected platforms to dam malicious microcode execution.
Some platforms additionally require a SEV firmware replace for SEV-SNP attestation, with customers having to replace the system BIOS and reboot to allow attestation of the mitigation.
To substantiate that the mitigation has been accurately put in, verify whether or not the microcode model(s) matches the one(s) listed within the desk under.
| Code Title | Household | CPUID |
| Naples | AMD EPYC 7001 Sequence | 0x00800F12 |
| Rome | AMD EPYC 7002 Sequence | 0x00830F10 |
| Milan | AMD EPYC 7003 Sequence | 0x00A00F11 |
| Milan-X | AMD EPYC 7003 Sequence | 0x00A00F12 |
| Genoa | AMD EPYC 9004 Sequence | 0x00A10F11 |
| Genoa-X | AMD EPYC 9004 Sequence | 0x00A10F12 |
| Bergamo/Siena | AMD EPYC 9004 Sequence | 0x00AA0F02 |
“We’ve demonstrated the power to craft arbitrary malicious microcode patches on Zen 1 by Zen 4 CPUs. The vulnerability is that the CPU makes use of an insecure hash perform within the signature validation for microcode updates,” the Google Safety Group mentioned.
“This vulnerability might be utilized by an adversary to compromise confidential computing workloads protected by the most recent model of AMD Safe Encrypted Virtualization, SEV-SNP or to compromise Dynamic Root of Belief Measurement.”
Google security researchers, credited with discovering and reporting this flaw to AMD, have additionally shared a proof-of-concept (PoC) exploit (examined on AMD EPYC and AMD Ryzen 9 CPUs) that reveals how attackers can create arbitrary microcode patches.
Their PoC exploit makes the RDRAND instruction on susceptible AMD Zen processors all the time return 4, which additionally units the carry flag (CF) to 0. This means that the return worth is invalid and ensures the exploit cannot be used “to compromise accurately functioning confidential computing workloads.”
This week, AMD has additionally acquired a report from Li-Chung Chiang at NTU (Nationwide Taiwan College) detailing cache-based side-channel assaults towards Safe Encrypted Virtualization (SEV) that affect information heart (1st Gen to 4th Gen AMD EPYC) and embedded (AMD EPYC 3000/7002/7003/9004) processors.
AMD suggested builders to comply with greatest practices for prime and probe assaults (e.g., constant-time algorithms), keep away from secret-dependent information each time attainable, and comply with the steering concerning Spectre-type assaults.
