A specially-crafted file can cause libjxl's decoder to write pixel data to uninitialized unallocated memory. Soon after that data from another uninitialized unallocated region is copied to pixel data.

This can be done by requesting color transformation of grayscale images to another grayscale color space. Buffers allocated for 1-float-per-pixel are used as if they are allocated for 3-float-per-pixel. That happens only if LCMS2 is used as CMS engine. There is another CMS engine available (selected by build flags).

An out of bounds read exists in libjxl. An attacker using a specifically crafted file could cause an out of bounds read in the exif handler. We recommend upgrading to version 0.8.1 or past commit  https://github.com/libjxl/libjxl/pull/2101/commits/d95b050c1822a5b1ede9e0dc937e43fca1b10159 https://github.com/libjxl/libjxl/pull/2101/commits/d95b050c1822a5b1ede9e0dc937e43fca1b10159

An issue was discovered in decpatchdictionary.cc in libjxl before 0.8.2. An integer underflow in patch decoding can lead to a denial of service, such as an infinite loop.

A specially-crafted file can cause libjxl's decoder to read pixel data from uninitialized (but allocated) memory.

This can be done by causing the decoder to reference an outside-image-bound area in a subsequent patches. An incorrect optimization causes the decoder to omit populating those areas.

For certain valid JPEG XL images with a size slightly larger than an integer number of groups (256x256 pixels) when processing the groups out of order the decoder can perform an out of bounds copy of image pixels from an image buffer in the heap to another. This copy can occur when processing the right or bottom edges of the image, but only when groups are processed in certain order. Groups can be processed out of order in multi-threaded decoding environments with heavy thread load but also with images that contain the groups in an arbitrary order in the file. It is recommended to upgrade past 0.6.0 or patch with https://github.com/libjxl/libjxl/pull/775