GIMP CVE-2026-59091

Exploitation requires a user to actively open a specially crafted PSD or PAA image file in GIMP - the out-of-bounds write is triggered during file parsing by the respective plugin. No network service, daemon, or open port is involved; GIMP must be executed in an interactive session by a local user (AV:L, UI:R). The attacker must successfully deliver the malicious file and convince the target to open it, typically via phishing, malicious download links, or shared storage. What limits exploitation: the local-only attack surface eliminates remote scanning or automated mass exploitation (SSVC: not automatable); the EPSS score of 0.17% confirms low observed exploitation activity; no authentication bypass, privilege escalation beyond the current user context, or kernel access is implied (S:U, PR:N means any unprivileged user session suffices to be targeted). The vulnerability is not triggered by background file indexing or thumbnail generation - GIMP must be the processing application.

CVSS 7.8 with a High rating is justified by the C:H/I:H/A:H impact triad, but the local attack vector (AV:L) and required user interaction (UI:R) meaningfully constrain real-world exploitation. The EPSS score of 0.17% (6th percentile) is very low, indicating the security community sees minimal current exploitation probability - consistent with SSVC classifying exploitation as none and the attack as not automatable. No KEV listing and no identified POC further reduce urgency relative to the raw CVSS score. The SSVC technical impact of total does confirm that successful exploitation yields full compromise of the affected user session. On balance, this is a genuine security issue warranting patching in the normal patch cycle, but it does not represent an imminent emergency: it requires targeted delivery, user action, and does not propagate without human involvement.

An attacker emails or hosts a maliciously crafted PSD or PAA image file, then social-engineers the target into opening it with GIMP. The malformed image triggers an out-of-bounds write in the file format plugin, corrupting heap or stack memory in a way that redirects control flow, ultimately executing attacker-supplied shellcode under the victim's user account. No public proof-of-concept has been identified at time of analysis, but the attack requires no special privileges and low technical complexity once the user opens the file.