Intel QuickAssist Technology CVE-2025-35998

Exploitation requires local access on an Intel platform that has QuickAssist Technology present and active, and the attacker must already hold privileged, Ring 0 / kernel-level access (PR:H) - a 'system software adversary with a privileged user.' The CVSS 4.0 vector additionally sets AT:P (attack requirement present), and the description specifies 'special internal knowledge' of the device is needed, meaning successful abuse depends on detailed understanding of the alternate hardware interface. No user interaction is required (UI:N). Limiting factors: it is not remotely reachable (AV:L), it does not work without pre-existing high privileges, it applies only to QAT-enabled Intel platforms (not default on commodity desktops/laptops without QAT), and the attack-requirement plus internal-knowledge prerequisites narrow it toward capable, targeted adversaries rather than opportunistic attackers.

Signals are consistent and point to a real-but-niche, low-urgency issue rather than a mass-exploitation threat. The CVSS 4.0 vector (AV:L/AC:L/AT:P/PR:H/UI:N/VC:H/VI:H/VA:N, base 7.0) requires local access and already-high privileges, with an explicit attack requirement (AT:P) and the description's note that 'special internal knowledge' is needed - the attacker must essentially already be a kernel/system-software adversary. EPSS is 0.01% (0th percentile), there is no public exploit identified at time of analysis, and the CVE is not in CISA KEV, so widespread or opportunistic exploitation is not indicated. The high PR:H requirement substantially limits practical reach: this is a privilege-boundary/defense-in-depth hardening fix (escaping certain kernel confinement or accessing protected device state) rather than a remote or low-privilege foothold. Note one tension to verify with the vendor: the CVSS marks AC:L (low complexity) while the description simultaneously cites attack requirements and 'special internal knowledge,' which in practice raises the real-world bar above what AC:L alone suggests.

An adversary who already controls privileged, kernel-level (Ring 0) code on a QAT-enabled host - for example via a malicious or compromised driver, or a confined system-software component seeking to break out - invokes the unprotected alternate hardware interface that lacks the protection enforced on the primary path. Leveraging special internal knowledge of the device, they read or tamper with protected kernel/device state to elevate privileges or breach confinement. No public proof-of-concept is identified at time of analysis, and the local, high-privilege prerequisites make this a low-complexity move only for an actor already operating at that trust level.