Parallels Desktop CVE-2024-54189

This vulnerability presents a HIGH real-world risk despite moderate CVSS scoring constraints. Risk signals include: (1) CVSS 7.8 with AV:L/AC:L/PR:L indicates LOCAL attack vector with LOW complexity and LOW privilege requirements—a normal user can trigger this without special access or user interaction; (2) The attack is deterministic and not dependent on timing windows for typical exploitation scenarios; (3) CWE-62 vulnerabilities are well-understood by attackers and frequently exploited in privilege escalation chains; (4) The impact is CRITICAL from a confidentiality, integrity, and availability perspective (C:H/I:H/A:H)—successful exploitation grants root-level file write access. EPSS probability is likely elevated given the straightforward exploitation path. Without KEV confirmation or POC public disclosure data in references, assume this is either pre-disclosure or early-stage exploitation. The attack surface is moderate (requires local account on Mac running Parallels Desktop), but Parallels Desktop is widely deployed in enterprise and development environments. This should be prioritized as CRITICAL for affected organizations, particularly those running version 20.1.1 (build 55740).

An attacker with a standard user account on a Mac running Parallels Desktop 20.1.1 (build 55740) creates a hard link from a sensitive file (e.g., ~/.ssh/authorized_keys or a LaunchDaemon plist in /Library/LaunchDaemons) to a temporary location that the snapshot process will write to. The attacker then initiates a VM snapshot operation, which triggers the root-level service to write data to the hard-linked target. Due to improper file ownership validation (CWE-62), the root process follows the hard link and writes to the sensitive system file. The attacker can then: (1) inject SSH keys into authorized_keys for remote code execution as the user, (2) plant a malicious LaunchDaemon to execute code at boot as root, or (3) modify shell initialization files to achieve persistent privilege escalation. The entire attack chain is non-interactive (UI:N per CVSS) and requires only standard user permissions (PR:L), making it exploitable by any local user. No public POC availability is noted, but the attack mechanics are straightforward and likely to be weaponized post-disclosure.