Sophos Endpoint CVE-2026-18367

Exploitation requires local access to a macOS system running a vulnerable build - Sophos Endpoint for macOS older than 2026.1.1 or Sophos Home for macOS older than 10.11.6 - with the Sophos security agent installed and running (its privileged helper/daemon must be present, which is the default for any protected endpoint). Per the CVSS vector no authentication or privileges are needed beyond the ability to run code locally (PR:N, UI:N), and no special product configuration is called out in the description. The primary limiting factor is the local-access requirement (AV:L): a purely remote, unauthenticated attacker cannot trigger this directly and must first gain a local foothold; hosts already patched to the fixed versions or not running the Sophos macOS agent are not exploitable.

The CVSS 3.1 vector (AV:L/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H, base 9.3) describes a low-complexity local attack requiring no privileges and no user interaction, with a scope change reflecting escalation to root - an unusually severe LPE profile. Real-world risk is bounded by the local-access prerequisite: an attacker must already have some presence on the macOS host (a standard user account, a phished session, or a foothold from another bug), making this a powerful post-exploitation and lateral-movement primitive rather than a remote entry point. No EPSS score, KEV listing, or POC was provided, so there is no evidence of active exploitation and no public exploit identified at time of analysis. The vendor tags 'RCE' and 'Authentication Bypass' appear to overstate the vector, since the CVSS is explicitly local (AV:L) and the mechanism is an authorization bypass rather than network RCE - rely on the AV:L vector over the tag labels.

An attacker who has obtained a low-privileged local account on a macOS host - for example through phishing, a stolen laptop, or a foothold from a separate application bug - invokes the improperly authorized privileged interface of the Sophos agent and executes their own code as root. No user interaction or additional credentials are required (AV:L/AC:L/PR:N/UI:N), and while no public exploit is identified at this time, the low attack complexity suggests reliable exploitation is feasible once the interface is understood.