Ghidra EUVD-2026-36008

Exploitation requires the victim to be running Ghidra on Windows (the cmd.exe metacharacter escaping flaw does not apply to the Linux or macOS launch paths), to open a Ghidra project, archive, or comment payload supplied by the attacker, and to actively click a URL annotation embedded in a program comment (CVSS UI:A confirms required user interaction). No authentication to Ghidra is required (PR:N) and no special non-default Ghidra configuration is needed - URL annotations are a standard listing feature. Limiting factors: the attacker must socially engineer the victim into loading attacker-controlled annotation content and clicking it, and code execution is bounded to the privileges of the Ghidra user, so analysts who run Ghidra under a low-privilege account or inside a sandbox materially reduce blast radius.

The CVSS 4.0 vector (AV:L/AC:L/AT:N/PR:N/UI:A) describes a local-vector, low-complexity, no-privileges-required attack that depends on active user interaction, yielding a base score of 8.4 with high impact across confidentiality, integrity, and availability of the vulnerable component. There is no EPSS score, KEV listing, or SSVC decision data in the provided input, and no public proof-of-concept identified at time of analysis, so real-world exploitation pressure is currently low despite the high CVSS. The vector labels AV:L because exploitation triggers from a file/project the victim opens locally, but the delivery is effectively social - a poisoned Ghidra project, archive, or shared comment data from an untrusted source. Practical risk is concentrated in reverse-engineering teams who routinely open third-party Ghidra archives, sample sharing platforms, or CTF artifacts on Windows; for those users this is a real priority because executing arbitrary code as the analyst frequently means access to malware samples, internal tooling, and lab credentials. For air-gapped or Linux/macOS Ghidra users, real risk is materially lower since the bug is scoped to the Windows cmd.exe launch path.

An attacker shares a Ghidra project archive, exported listing, or comment snippet (e.g. via a malware-sharing community, a tutorial, or a CTF challenge) containing a URL annotation whose target embeds cmd.exe metacharacters such as 'http://example.com/ & powershell -enc ...'. A Windows-based analyst opens the project and clicks the URL annotation expecting browser navigation, at which point cmd.exe interprets the appended command and executes it under the analyst's account. No public exploit is identified at time of analysis, but the bug class (argument injection via cmd.exe) is well understood and trivial to weaponize once the exact unescaped sink is known.