VMware ESXi CVE-2025-62624

Exploitation requires the host to be an ESXi 8.x or 9.x system with AMD-Pensando DPU hardware installed and the ionic driver loaded - hosts without this DPU are not exploitable. The attacker must already hold local, low-privileged access to the host (CVSS AV:L/PR:L), so this is a local privilege-escalation primitive, not remote unauthenticated compromise. Exploitation is further limited by high attack complexity (AC:H), meaning the attacker must satisfy heap-grooming/timing conditions that are not reliably reproducible, consistent with SSVC Automatable: no. No user interaction is required (UI:N). These preconditions - DPU hardware presence, prior local privilege, and high complexity - substantially narrow realistic exploitation despite the total technical impact.

The signals are internally consistent toward a serious-but-hard-to-exploit issue. The CVSS 4.0 vector (AV:L/AC:H/AT:N/PR:L/UI:N) shows exploitation requires local access and existing low privileges on the host, plus high attack complexity - likely reflecting the difficulty of grooming the heap and the timing/conditions needed to control an overflow in a kernel driver. Against that, the impact is total (VC:H/VI:H/VA:H and subsequent-system SC:H/SI:H/SA:H; SSVC Technical Impact: total), consistent with full hypervisor compromise. Exploitation-likelihood signals are all low: EPSS is 0.01% (3rd percentile), SSVC Exploitation is 'none', SSVC Automatable is 'no', there is no KEV listing, and no public exploit identified at time of analysis. The realistic priority is therefore moderate: high consequence if exploited, but low probability and meaningful preconditions (local low-priv foothold, DPU hardware present, high complexity). This is not a drop-everything internet-facing emergency; it is a patch-in-normal-cycle item for the specific subset of hosts running AMD-Pensando DPUs. One caveat: the CVSS 4.0 vector implies AV:L/PR:L but the AMD bulletin should be verified for the exact local privilege context (guest VM vs. host management plane), as that materially changes the threat model.

An attacker who has already gained low-privileged local execution on an ESXi 8.x/9.x host with an AMD-Pensando DPU (for example via a compromised management account or an initial foothold) sends crafted input to the ionic driver to overflow a heap buffer, grooming memory to overwrite kernel data and escalate to full host code execution. The high attack complexity means multiple attempts and specific heap conditions are likely required, and no public exploit is identified at time of analysis, making this a plausible post-compromise escalation rather than an initial-access vector.