GIMP CVE-2026-18306

Exploitation requires the target to actively open a crafted SGI raster image file in GIMP 3.2.2 on any supported platform (Windows, Linux, macOS). The CVSS vector PR:N confirms no attacker-side privileges are required; UI:R confirms the victim must perform an explicit file-open action. There is no network-listening attack surface - GIMP does not expose a network service, so the attacker's delivery mechanism is entirely social engineering (phishing, malicious download, shared network drive). Exploitation is bounded by S:U (scope unchanged), meaning code runs at the victim user's privilege level with no implied escalation to root or SYSTEM. No non-default configuration or special feature enablement is required; GIMP 3.2.2's default SGI parser is the vulnerable component.

The CVSS 7.8 score with vector AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H accurately reflects the attack model: 'local' here means a file-open action by the victim rather than physical machine access, no attacker privileges are required, but successful exploitation is gated on user interaction. The full C:H/I:H/A:H triad maps to complete process compromise. No EPSS score was provided in the input, preventing quantitative exploitation-probability comparison; the absence of a KEV entry suggests no confirmed mass or targeted exploitation at time of analysis. ZDI's public advisory (ZDI-26-459) significantly increases post-disclosure exploitation risk, as it provides technical detail sufficient for exploit development - even a mid-sized development or design team should treat this as urgent given GIMP's prevalence in creative and technical workflows. The patch commit on GNOME GitLab is confirmed, but the absence of a formally released patched build means users are currently dependent on building from source or waiting for distribution packages. No conflicting signals between the CVSS vector and the description were identified; the local file-open trigger is consistent with AV:L/UI:R.

An attacker crafts a malicious SGI image file with manipulated header fields designed to trigger an integer overflow in GIMP's parser, then delivers it via a phishing email or a drive-by download page. The victim opens the file in GIMP 3.2.2, the parser overflows during dimension or channel calculation, and the resulting heap corruption enables the attacker to redirect execution to shellcode running as the victim's local user account. No public proof-of-concept exploit code has been identified at time of analysis, but ZDI's published advisory provides sufficient technical detail to guide exploit development.