AMD Platform Management Framework CVE-2025-29937

Exploitation requires local access to the system with authenticated user privileges (PR:L per CVSS). The attacker must be able to execute code or issue commands as a valid local user account. The vulnerability is not exploitable remotely or without authentication. Specifically, the out-of-bounds read must be triggered via an interface accessible to the PMF (likely a driver ioctl, power management API, or thermal sensor interface), which may require the attacker to be a member of a specific group (e.g., 'root', 'wheel', or a power management group) depending on system configuration. Default configurations of affected Ryzen systems are vulnerable if a local attacker has standard user-level access; however, tightened local access controls, read-only filesystem enforcement, or driver access restrictions can mitigate exposure. The vulnerability does not require user interaction (UI:N) once the attacker has local code execution.

CVSS 5.8 reflects local attack vector (AV:L), low complexity (AC:L), and low privilege requirement (PR:L), meaning an authenticated local user can trigger exploitation. The CVSS impact vector shows high availability risk (VA:H) and low confidentiality risk (VC:L), suggesting memory disclosure or potential denial of service via targeted reads. The AT:P (Temporal: Prepare) modifier indicates some timing or environmental preparation may be required, reducing real-world ease of exploitation compared to AC:L alone. No EPSS score, KEV listing, or public POC is documented in the input, classifying this as a reported vulnerability without confirmed active exploitation. Risk is moderate for systems where local user containment is weak or where the PMF manages sensitive processor state; however, the local-only attack vector significantly limits exposure compared to remote vulnerabilities. Organizations with tight local access controls or containerized/virtualized deployments face lower practical risk. The confidentiality impact (VC:L) suggests limited but real information disclosure rather than critical secret exfiltration.

A local, authenticated user on an affected AMD Ryzen system executes a crafted utility or script that triggers an out-of-bounds read in the Platform Management Framework. This read accesses memory beyond the intended buffer, potentially exposing kernel memory, firmware state, or other sensitive data to the attacker's user-space process. The attacker may repeat this to map sensitive memory regions, extract cryptographic keys, or gather information for further privilege escalation attacks.