Google Chrome CVE-2026-106384

Hard prerequisite: the attacker must have ALREADY compromised a Chrome renderer process (e.g., via a prior renderer RCE), because the flaw lets that already-controlled renderer bypass Site Isolation boundaries using a crafted HTML page. Site Isolation is enabled by default, so no special configuration is needed - but exploitation is strictly limited to Chrome versions before 155.0.8059.39 and cannot occur without the preceding renderer compromise, making this a chained, second-stage vulnerability rather than a standalone remote attack. The vendor vector also marks UI:R (user must load a page), though a compromised renderer typically already controls page content.

This is a genuine-but-moderate priority, not a drop-everything issue. The signals are consistent and point the same direction: Chromium rated it Medium severity, the CVSS base score is 6.5 (CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:N), EPSS is very low at 0.19% (8th percentile), it is NOT listed in CISA KEV, and no public exploit has been identified. The single most important mitigating factor is baked into the description but UNDER-REPRESENTED in the vendor CVSS vector: exploitation requires an attacker to have ALREADY compromised the renderer process. That makes this a second-stage/sandbox-escape-adjacent bug, not an independently exploitable network flaw. CONFLICT TO FLAG: the vendor vector encodes PR:N (no privileges), yet the description explicitly requires a compromised renderer - a substantial prerequisite that in my independent assessment maps to PR:H. Also note the impact labeling: Google scored this I:H/C:N, but a Site Isolation bypass conceptually enables cross-site data access, which reads more like a confidentiality impact - verify intended impact with the vendor. Net: real defect worth patching on normal Chrome update cadence, but the compromised-renderer precondition and low EPSS argue strongly against treating it as an urgent, widely-exploitable threat.

An attacker first gains code execution inside a Chrome renderer process through a separate vulnerability, then serves a crafted HTML page that abuses the Site Isolation weakness to reach content or data belonging to a different site that should have been isolated. No public exploit code has been identified, and the attack is only meaningful as the second stage of a chain that already requires renderer compromise.