Microsoft XPS CVE-2026-68891

Exploitation requires a local session with at minimum low-privileged user credentials (PR:L per CVSS). The attack additionally demands high complexity (AC:H), meaning non-default or difficult-to-reproduce conditions must be met - likely involving a specifically crafted XPS document that triggers the parser's out-of-bounds read. The vulnerability is limited to information disclosure (C:H, I:N, A:N); no code execution or integrity impact applies. Exploitation is bounded to the local system (AV:L, S:U), ruling out remote or network-based attack paths. The XPS processing component must be invoked, either directly or via a Windows subsystem that parses XPS content.

Real-world risk is moderate despite the C:H confidentiality rating. The CVSS vector AV:L/AC:H/PR:L/UI:N/S:U constrains exploitation to local sessions with low-privilege accounts and requires high attack complexity - meaning reliable triggering is non-trivial. EPSS data was not provided; given the local-only, high-complexity profile and absence of KEV listing or POC code, automated or widespread exploitation is unlikely. The primary concern is targeted post-compromise or insider-threat scenarios where an attacker already on the box attempts to extract memory contents (e.g., credentials, cryptographic material) via crafted XPS documents. The breadth of affected versions (every supported and some legacy Windows SKUs) makes the patch surface large, but the attack constraints keep this below critical priority for most organizations. No exploitability signals (KEV, POC, EPSS) contradict the CVSS score here - this reads as a genuine moderate finding.

A local user with a standard Windows account plants a specially crafted XPS document and triggers its processing through the XPS stack - for example, via a print-to-XPS workflow or document preview - under conditions that satisfy the high-complexity trigger. The out-of-bounds read exposes memory adjacent to the XPS parser's buffer, potentially leaking sensitive data such as heap contents, tokens, or cryptographic material accessible to the process. No public exploit code exists at time of analysis, making this scenario primarily relevant to advanced post-compromise or insider threat contexts.