Google Chrome CVE-2026-78967

Exploitation requires two conditions to be simultaneously met: first, the attacker must have already compromised the Chrome renderer process via a separate vulnerability - this is not a standalone remote-code-execution flaw. Second, the victim must interact with attacker-controlled content (UI:R per CVSS), typically by visiting a crafted HTML page. Exploitation is limited to Chrome versions prior to 152.0.7977.65. The renderer-compromise prerequisite is the primary limiting factor: an attacker without a working renderer exploit cannot leverage this vulnerability. There is no indication that non-default configurations are required beyond standard Chrome operation with BFCache enabled, which is the default.

The provided CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:N) scores 6.5 Medium, but this requires qualification: the description explicitly states the attacker must have already compromised the renderer process, which represents a significant prerequisite not reflected in the AC:L metric. In practice, exploiting this vulnerability requires chaining it with a separate renderer compromise vulnerability, substantially raising real-world attack complexity beyond what AC:L implies - this discrepancy should be flagged when triaging. EPSS at 0.38% (31st percentile) confirms low community-observed exploitation probability. The vulnerability is not in CISA KEV and no public POC has been identified. Despite the High integrity impact, the renderer-compromise prerequisite limits this to sophisticated, multi-stage attacks rather than opportunistic exploitation. The effective real-world risk is lower than the CVSS score suggests in isolation.

An attacker first exploits a separate, unpatched Chrome renderer vulnerability to gain code execution within the sandboxed renderer process. They then serve a crafted HTML page that triggers BFCache navigation events, exploiting the missing authorization check to escalate from renderer-level access to system-level resource access, bypassing the Chrome sandbox boundary. No public exploit code has been identified at time of analysis, and this scenario requires chaining at least two vulnerabilities.