Python Pillow EUVD-2026-43730

The target application must accept and decode a JPEG2000 (.jp2/.j2k) image supplied by the attacker using an affected Pillow version (8.2.0-12.2.0), and the malicious file must be tiled so that per-tile widths accumulate - a single-tile image does not trigger the bug. No authentication, privileges, or user interaction are required (PR:N/UI:N) once such an input reaches the decoder. Exploitation is limited to deployments that actually process untrusted JPEG2000 input; services that only handle other formats, or that disable the JPEG2000 plugin, are not exposed. Impact is confined to availability (memory exhaustion / OOM) - no data disclosure or code execution.

This is a genuine but narrow-impact availability issue rather than a code-execution threat. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:N/UI:N with VC:N/VI:N/VA:H) correctly reflects an unauthenticated, network-reachable, availability-only impact and no confidentiality or integrity effect - consistent with a memory-exhaustion DoS and NOT with the supplied 'Information Disclosure' tag, which appears mislabeled given VC:N. EPSS was not provided and the CVE is not in CISA KEV; with no public exploit identified, real-world exploitation is currently theoretical, though the public patch and diff lower the barrier to weaponization. Practical risk is highest for services that auto-decode user-supplied images (upload pipelines, thumbnailers, image proxies) and is effectively nil for deployments that never process untrusted JPEG2000 input.

An attacker uploads a small, specially crafted tiled JPEG2000 file to a web service that automatically generates thumbnails or previews with Pillow. When Pillow decodes it, the accumulated tile-width miscalculation drives a massive transient allocation, exhausting memory and crashing the worker or host. No public exploit code has been identified, but the low attack complexity and unauthenticated network vector make abuse straightforward once a payload is constructed from the public patch.