CVE-2026-96644

Pillow versions 11.0.0 through 13.0.0 mishandle multi-frame images (PSD, MPO, DCX, MIC). When a later frame switches to a wider mode (RGB/CMYK), the stale narrow-mode buffer is reused, causing out-of-bounds writes and crashes, and in some runtime profiles command execution.

Pillow, a Python imaging library, from 2.6.0 until 13.0.0 is vulnerable to argument injection when invoking Ghostscript for EPS input. EpsImagePlugin.Ghostscript passes a caller-controlled input filename after -f without converting it to a safe absolute path, so a name beginning with a hyphen can be interpreted as a Ghostscript option.

Pillow prior to 13.0.0 has a JPEG decoder vulnerability where a full-width scanline can be written into a buffer sized for a requested draft scale. ImagingJpegDecode does not verify libjpeg's actual output width after scaling, and an 8-bit lossless SOF3 JPEG decoded by libjpeg-turbo 3.x can ignore the requested downscaling geometry.

Pillow prior to 13.0.0 has a FlashPix JPEG-tile decoder vulnerability that can write decoded scanlines beyond the allocated tile buffer. FlashPix (FPX) JPEG decoding selects raw mode LA with two bytes per pixel, while an unrecognized JPEG mode can be decoded as JCS_UNKNOWN with four components, causing ImagingJpegDecode to write four-component scanlines into the two-byte tile buffer.

Pillow has a heap buffer overflow and stale-heap disclosure in _tiffWriteProc when saving compressed TIFF data through libtiff to a target without a usable fileno(), such as BytesIO. The callback sizes its allocation from the prior buffer size instead of the current write position, allowing data to be copied beyond the allocation. This is fixed in version 13.0.0.

A vulnerability in Pillow before version 12.3.0 allows an attacker to read process memory or cause a fault when loading an uncompressed McIdas AREA image via the mmap codec path. Attacker-controlled header words can set a row stride smaller than the natural row width, leading to out-of-bounds read beyond the mapped region.

Integer overflow in the ImagingResampleHorizontal function in Pillow before version 3.1.1 allows remote attackers to have unspecified impact via negative values of the new size, leading to a heap-based buffer overflow.

Pillow versions 11.2.1 through 13.0.0 incorrectly assume image rows are tightly contiguous when exporting to Arrow. When PILLOW_ALIGNMENT causes row padding or an aligned offset before the first row, the resulting Arrow buffer may include uninitialized padding, bytes preceding the first row, or stale heap contents from previously freed images. This can disclose one user's pixels or other process memory to another user.

Pillow is a Python imaging library. From 9.4.0 until 13.0.0, ImagingJpegEncode() in src/libImaging/JpegEncode.c can call jpeg_write_marker() for a JPEG COM comment when an ICC profile has exactly filled the 65536-byte encoder buffer. libjpeg writes one byte before checking the exhausted free_in_buffer counter, which then wraps and permits the attacker-controlled comment and subsequent compressed image data to overwrite adjacent heap memory. Re-saving an untrusted image as JPEG while preserving its ICC profile and forwarded comment can crash the process and may enable code execution in suitable runtime and heap layouts. This issue is fixed in version 13.0.0.

In Pillow from 11.0.0 until 13.0.0, Exif.load() in src/PIL/Image.py repeatedly slices the remaining EXIF data while it begins with the six-byte Exif header. Because src/PIL/PngImagePlugin.py accepts an uncapped PNG eXIf chunk, a file filled with repeated headers causes a new shrinking copy on every iteration and quadratic CPU work. The condition is reachable through Image.getexif(), Image._getexif(), and ImageOps.exif_transpose(), including common photo-upload rotation workflows. This issue is fixed in version 13.0.0.