OpenEXR Heap-Based Buffer Overflow in Deep Scanline Parsing via Forged Unpacked Size

Summary

The OpenEXRCore code is vulnerable to a heap-based buffer overflow during a write operation when decompressing ZIPS-packed deep scan-line EXR files with a maliciously forged chunk header.

Details

When parsing STORAGE_DEEP_SCANLINE chunks from an EXR file, the following code (from src/lib/OpenEXRCore/chunk.c) is used to extract the chunk information:

if (part->storage_mode == EXR_STORAGE_DEEP_SCANLINE)
// SNIP...
        cinfo->sample_count_data_offset = dataoff;
        cinfo->sample_count_table_size  = (uint64_t) ddata[0];
        cinfo->data_offset              = dataoff + (uint64_t) ddata[0];
        cinfo->packed_size              = (uint64_t) ddata[1];
        cinfo->unpacked_size            = (uint64_t) ddata[2];
// SNIP...

By storing this information, the code that will later decompress and reconstruct the chunk bytes, will know how much space the uncompressed data will occupy.

This size is carried along in the chain of decoding/decompression until the undo_zip_impl function in src/lib/OpenEXRCore/internal_zip.c:

static exr_result_t
undo_zip_impl (
    exr_decode_pipeline_t* decode,
    const void*            compressed_data,
    uint64_t               comp_buf_size,
    void*                  uncompressed_data,
    uint64_t               uncompressed_size,
    void*                  scratch_data,
    uint64_t               scratch_size)
{
    size_t       actual_out_bytes;
    exr_result_t res;

    if (scratch_size < uncompressed_size) return EXR_ERR_INVALID_ARGUMENT;

    res = exr_uncompress_buffer (
        decode->context,
        compressed_data,
        comp_buf_size,
        scratch_data,
        scratch_size,
        &actual_out_bytes);

    if (res == EXR_ERR_SUCCESS)
    {
        decode->bytes_decompressed = actual_out_bytes;
        if (comp_buf_size > actual_out_bytes)
            res = EXR_ERR_CORRUPT_CHUNK;
        else
            internal_zip_reconstruct_bytes (
                uncompressed_data, scratch_data, actual_out_bytes);
    }

    return res;
}

The uncompressed_size comes from the unpacked_size extracted earlier, and the uncompressed_data is a buffer allocated by making space for the size "advertised" in the chunk information.

However, scratch_data and actual_out_bytes will contain, after decompression, the uncompressed data and its size, respectively.

The vulnerability lies in the fact that the undo_zip_impl function lacks code to check whether actual_out_bytes is greater than uncompressed_size.