GIMP CVE-2026-58381

Exploitation requires a local victim to open a specially crafted Paint Shop Pro (.psp) format file in GIMP, invoking the PSP import parser and reaching read_layer_block(). It is not remotely triggerable without user action: the CVSS vector is AV:L with UI:R (required user interaction), so the attack depends on social engineering the target into opening the malicious image. No special GIMP configuration or elevated privileges are needed beyond the ability to run GIMP and open a file, which limits realistic impact to client-side, per-user compromise rather than server or unattended exploitation.

This is a genuine but low-urgency memory-corruption flaw. The Red Hat-provided CVSS 3.1 vector (AV:L/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:H, 7.3) is dominated by local, user-interaction-gated preconditions: the attacker cannot reach read_layer_block() over a network - a local user must open a specially crafted PSP (Paint Shop Pro) file in GIMP. All exploitation signals point away from active/widespread abuse: EPSS is 0.12% (2nd percentile), CISA SSVC records Exploitation=none, Automatable=no, and Technical Impact=partial, there is no KEV listing, and no public exploit was provided. The C:H/I:H/A:H rating reflects the theoretical worst case (arbitrary code execution from a double-free), but the description itself hedges ('potentially leading to... arbitrary code execution'), and SSVC's 'partial' technical impact suggests reliable code execution is not demonstrated - realistic outcome is a crash/DoS. One notable inconsistency to verify: the vendor assigns PR:L, yet a file-parsing bug typically needs no attacker privileges (the victim's action is already captured by UI:R), so PR:N/UI:R is the more defensible modeling and would slightly raise the base score. Net: real, patchable, but a routine-priority desktop client-side issue, not an emergency.

An attacker distributes a malicious .psp image (via email, download, or a shared file store) and social-engineers a victim into opening it in GIMP. When GIMP's PSP parser reaches read_layer_block(), the crafted layer data triggers a double-free (CWE-415), corrupting heap metadata and most likely crashing GIMP, with a possibility of code execution in the user's context. No public proof-of-concept exploit was provided, and exploitation requires local user interaction rather than any remote/network path.