Google Chrome CVE-2026-79112

Exploitation has two hard prerequisites: (1) the attacker must have already achieved code execution within the Chrome renderer process via a separate, independent vulnerability - this CVE is not independently exploitable from a clean browser state; (2) the victim must interact with a crafted HTML page (UI:R). The out-of-bounds read is confined to sandbox memory (S:U), so it cannot directly reach OS-level memory or other processes. Exploitation is therefore limited to chained attack scenarios where a renderer RCE vulnerability exists simultaneously. Non-default configurations are not required - any Chrome instance prior to 152.0.7977.65 with a compromised renderer is potentially affected. Authentication is not required on the target system (PR:N per CVSS), but the chained renderer compromise is the binding constraint.

Real-world risk here is meaningfully lower than the raw CVSS 6.5 suggests. The CVSS vector (AV:N/AC:L/PR:N/UI:R) presents this as a straightforward network-reachable flaw, but the description imposes a critical precondition absent from the vector: the attacker must have already compromised the renderer process via a separate vulnerability. This makes the actual attack complexity high, not low - the CVSS AC:L is misleading for operational prioritization. EPSS at 0.17% (7th percentile) reflects this reality: automated, widespread exploitation is unlikely. No KEV listing confirms no observed active exploitation. There is no confidentiality impact on systems outside the sandbox (S:U), limiting blast radius. Security teams should patch promptly as part of the normal Chrome update cycle but need not invoke emergency procedures; even a mid-sized IT operation should be able to absorb this through routine browser update policy rather than emergency patching.

An attacker first exploits a separate Chrome renderer vulnerability to gain code execution inside the renderer sandbox. They then serve a crafted HTML page containing malicious graphical content that triggers the Skia out-of-bounds read, leaking in-sandbox memory - potentially including heap layout data or partial credential fragments - back to the attacker. No public exploit code or POC has been identified at time of analysis.