Pillow CVE-2026-112397

Exploitation requires (1) the target application to decode an attacker-supplied PPM or PGM image that produces an image with width zero, and (2) the application to call Pillow's ImageMorph.MorphOp.apply() (a morphology operation such as erosion8/dilation) on that image. The limiting factor is the morphology code path: most Pillow deployments never invoke ImageMorph, so only applications explicitly performing morphological operations on untrusted images are affected. No authentication is required (PR:N) and no user interaction beyond supplying the image (UI:N); impact is confined to a process crash with no data exposure or code execution.

This is a denial-of-service memory-safety flaw with a narrow real-world risk profile despite its 8.7 (HIGH) CVSS 4.0 base score. The vector (CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H) correctly reflects the single meaningful consequence: availability impact only (VA:H) via a process crash, with no confidentiality or integrity impact (VC:N/VI:N). CWE-787 (out-of-bounds write) is listed, but the description and fix indicate an out-of-bounds memory access at outrow[width-1] when width is 0 - the demonstrated and fixed impact is a crash, not code execution, so the score's C:N/I:N is consistent. The network attack vector (AV:N) is realistic only indirectly: Pillow itself is a local library, so AV:N applies solely to applications that ingest attacker-supplied images over a network AND then invoke the morphology routine - a comparatively uncommon code path. There is NO EPSS score, NO CISA KEV listing, and NO public exploit (only a vendor fix commit and a test case are provided), so all active-exploitation and probability signals are absent rather than low. Net: this is a legitimate but low-urgency DoS to patch in the normal cycle for anyone using ImageMorph; it is not a high-real-risk emergency, and the HIGH severity reflects easy triggering conditions rather than severe impact. Note the signal tension: AV:N raises the paper score, but exploitability is gated by the application calling MorphOp.apply() on untrusted input - verify whether your deployment actually uses morphology operations.

An application accepts user-uploaded images and passes them through Pillow's ImageMorph.MorphOp.apply() (e.g., for preprocessing or OCR cleanup). An attacker uploads a crafted PPM or PGM file decoded to a zero-width image; the native morphology routine indexes outrow[width-1], reads/writes out of bounds, and crashes the worker process, causing denial of service. No public exploit code is provided - only the vendor fix commit and a regression test demonstrate the condition.