{"id":"CVE-2025-48074","aliases":["GHSA-x22w-82jp-8rvf","PYSEC-2026-1749"],"title":"OpenEXR Out-Of-Memory via Unbounded File Header Values","summary":"OpenEXR Out-Of-Memory via Unbounded File Header Values","severity":"medium","vendor":"openexr","product":"openexr","ecosystem":"pip","affected":["openexr >= 3.3.2, < 3.3.3"],"patched":["openexr 3.3.3"],"published":"2025-07-31","updated":"2026-07-07","source":"OSV","sourceUrl":"https://osv.dev/vulnerability/GHSA-x22w-82jp-8rvf","references":[{"url":"https://github.com/AcademySoftwareFoundation/openexr/security/advisories/GHSA-x22w-82jp-8rvf"},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2025-48074"},{"url":"https://github.com/AcademySoftwareFoundation/openexr/commit/2df7802a6421bf39ed73a607392f01b3c6da4e4a"},{"url":"https://github.com/AcademySoftwareFoundation/openexr"},{"url":"https://github.com/ShielderSec/poc/tree/main/CVE-2025-48074"}],"tags":["osv","pip"],"epss":0.00259,"epssPercentile":0.17962,"ingestedAt":"2026-07-08T18:25:53.945Z","slug":"CVE-2025-48074","body":"## Overview\n\n### Summary\nThe OpenEXR file format defines many information about the final image inside of the file header, such as the size of data/display window.\n\nThe application trusts the value of `dataWindow` size provided in the header of the input file, and performs computations based on this value.\n\nThis may result in unintended behaviors, such as excessively large number of iterations and/or huge memory allocations.\n\n\n### Details\nA concrete example of this issue is present in the function `readScanline()` in `ImfCheckFile.cpp` at line 235, that performs a for-loop using the `dataWindow min.y` and `max.y` coordinates that can be arbitrarily large.\n\n```cpp\nin.setFrameBuffer (i);\n\nint step = 1;\n\n//\n// try reading scanlines. Continue reading scanlines\n// even if an exception is encountered\n//\nfor (int y = dw.min.y; y <= dw.max.y; y += step) // <-- THIS LOOP IS EXCESSIVE BECAUSE OF DW.MAX\n{\n    try\n    {\n        in.readPixels (y);\n    }\n    catch (...)\n    {\n        threw = true;\n\n        //\n        // in reduceTime mode, fail immediately - the file is corrupt\n        //\n        if (reduceTime) { return threw; }\n    }\n}\n```\n\nAnother example occurs in the `EnvmapImage::resize` function that in turn calls `Array2D<T>::resizeEraseUnsafe` passing the `dataWindow` X and Y coordinates and perform a huge allocation.\n\nOn some system, the allocator will simply return `std::bad_alloc` and crash. On other systems such as macOS, the allocator will happily continue with a \"small\" pre-allocation and allocate further memory whenever it is accessed.\nThis is the case with the `EnvmapImage::clear` function that is called right after and fills the image RGB values with zeros, allocating tens of Gigabytes.\n\n### PoC\n\nNOTE: please download the `oom_crash.exr` file via the following link:\n \nhttps://github.com/ShielderSec/poc/tree/main/CVE-2025-48074\n\n1. Compile the `exrcheck` binary in a macOS or GNU/Linux machine with ASAN.\n2. Open the `oom_crash.exr` file with the following command:\n\n```\nexrcheck oom_crash.exr\n```\n\n3. Notice that `exrenvmap`/`exrcheck` crashes with ASAN stack-trace.\n\n### Impact\nAn attacker could cause a denial of service by stalling the application or exhaust memory by stalling the application in a loop which contains a memory leakage.\n\n## Affected packages\n\n- `openexr >= 3.3.2, < 3.3.3`\n\n## Remediation\n\nUpgrade to a patched release:\n\n- `openexr 3.3.3`","depth":"sunlit","depthScore":28,"depthScoreParts":{"impact":27.5,"likelihood":0.1,"exploitation":0,"ransomware":0},"changes":[]}