CVE-2026-65970

In OpenImageIO prior to versions 3.0.18.0 and 3.1.13.0, a signed 32-bit integer overflow in the SwapRGBABytes() function causes a large negative pointer offset when processing large kABGR DPX images. This results in immediate crashes and out-of-bounds reads and writes.

In OpenImageIO prior to versions 3.0.18.0 and 3.1.13.0, a signed 32-bit integer overflow in the pixel-loop index expression causes an out-of-bounds write that crashes the process.

In OpenImageIO versions prior to 3.0.18.0 and 3.1.13.0, a heap-based buffer overflow occurs due to a signed integer overflow in the QueryRGBBufferSizeInternal() function. This issue arises when processing crafted DPX image files, leading to writes outside the allocated buffer.

OpenImageIO prior to versions 3.0.18.0 and 3.1.13.0 contains a heap-based buffer overflow in the HEIF decoder, allowing out-of-bounds writes. This can lead to memory corruption and potential code execution.

OpenImageIO prior to versions 3.0.18.0 and 3.1.13.0 has a vulnerability related to buffer size calculation in jpeg2000input.cpp, leading to heap overflow. This issue occurs when the product of image dimensions exceeds the maximum value for 32-bit arithmetic, resulting in the allocation of an undersized buffer.

OpenImageIO prior to versions 3.0.18.0 and 3.1.13.0 contains a vulnerability that allows for heap overflow up to 65535 bytes due to improper clamping of run length in .pic files. The issue occurs in the code responsible for processing RLE.

OpenImageIO prior to versions 3.0.18.0 and 3.1.13.0 contains a vulnerability in sgiinput.cpp where the use of OIIO_DASSERT for bounds checking in the RLE decode loop does not function in release builds. As a result, a crafted .sgi file with RLE count exceeding scanline width causes heap buffer overflow and crash.

OpenImageIO is a toolset for reading, writing, and manipulating image files. Prior to 3.1.16.0, a crafted 1-bit contiguous cmyk tiff is exposed through a native uint1 imagespec, so callers allocate a bit-packed buffer. tiffinput::read_native_scanline_locked() nevertheless invokes tiffinput::bit_convert() with 8-bit output and writes one expanded byte per value into that smaller buffer, resulting in a heap out-of-bounds write and memory corruption. This issue is fixed in 3.1.16.0.

In OpenImageIO prior to 3.0.21.0, 3.1.16.0, and 3.2.0.3-beta1, a truncated TGA can leave a pending GIF frame that is processed during output close. gifsplitpalette() computes numpixels multiplied by the palette partition width in signed 32-bit arithmetic; a large image overflows that intermediate, corrupts subpixelsa, and drives an out-of-bounds read while building the GIF palette.

OpenImageIO is a toolset for reading, writing, and manipulating image files. Prior to 3.0.21.0, 3.1.16.0, and 3.2.0.3-beta1, a crafted cineon image can declare unsupported component bit depth 26. cineoninput::open() maps it to a 32-bit imagespec, but libcineon maps the unsupported depth to an 8-byte value, so cineoninput::read_native_scanline() causes attacker-controlled data to be written beyond the 4-byte-per-pixel caller buffer, resulting in a heap out-of-bounds write and memory corruption. This issue is fixed in versions 3.0.21.0, 3.1.16.0, and 3.2.0.3-beta1.