GHSA-cjc8-g9w8-chfw imagemagick: heap-buffer overflow read in MNG magnification with alpha

• NuGet Magick.NET-Q16-AnyCPU: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-HDRI-AnyCPU: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-HDRI-OpenMP-arm64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-HDRI-OpenMP-x64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-HDRI-arm64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-HDRI-x64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-HDRI-x86: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-OpenMP-arm64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-OpenMP-x64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-arm64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-x64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q16-x86: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q8-AnyCPU: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q8-OpenMP-arm64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q8-OpenMP-x64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q8-arm64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q8-x64: < 14.8.0 → fixed in 14.8.0 • NuGet Magick.NET-Q8-x86: < 14.8.0 → fixed in 14.8.0

Vulnerability Details When performing image magnification in `ReadOneMNGIMage` (in `coders/png.c`), there is an issue around the handling of images with separate alpha channels. When loading an image with a color type that implies a separate alpha channel (ie. `jng_color_type >= 12`), we will load the alpha pixels in this loop: ```c if (logging != MagickFalse) (void) LogMagickEvent(CoderEvent,GetMagickModule(), " Reading alpha from alpha_blob."); jng_image=ReadImage(alpha_image_info,exception); if (jng_image != (Image *) NULL) for (y=0; y rows; y++) { s=GetVirtualPixels(jng_image,0,y,image->columns,1,exception); q=GetAuthenticPixels(image,0,y,image->columns,1,exception); // [0] if ((s == (const Quantum *) NULL) || (q == (Quantum *) NULL)) break; if (image->alpha_trait != UndefinedPixelTrait) for (x=(ssize_t) image->columns; x != 0; x--) { SetPixelAlpha(image,GetPixelRed(jng_image,s),q); q+=(ptrdiff_t) GetPixelChannels(image); s+=(ptrdiff_t) GetPixelChannels(jng_image); } else for (x=(ssize_t) image->columns; x != 0; x--) { Quantum alpha; alpha=GetPixelRed(jng_image,s); SetPixelAlpha(image,alpha,q); if (alpha != OpaqueAlpha) image->alpha_trait=BlendPixelTrait; // [1] q+=(ptrdiff_t) GetPixelChannels(image); s+=(ptrdiff_t) GetPixelChannels(jng_image); } if (SyncAuthenticPixels(image,exception) == MagickFalse) break; } ``` Note that at \[1\] we update `image->alpha_trait`, but if our alpha image only contains non-opaque pixels in the last row, we do not call `GetAuthenticPixels` (at \[0\]) after this change has been made. The next call to `GetAuthenticPixels` will then call down into `ResetPixelChannelMap` which adds the new alpha channel to the image channel mappings and metadata. If we then pass this image into the `MAGN` chunk type, we can see that at \[2\] we calculate the sizes for intermediate buffers `next` and `prev`, before calling `GetAuthenticPixels` at \[4\]. After the call at \[4\], the `image->num_channels` has increased to include the new alpha channel, and now `length` and the previously allocated `next` and `prev` buffers are too small. Fortunately `length` is always used when copying into the buffers, but when reading pixels from the buffers, we call `GetPixelXXX` which assumes the layout of the current image, which requires a larger allocation. The pixel copying loop will subsequently read beyond the end of the allocation at \[5\]. ```c /* magnify the rows into the right side of the large image */ if (logging != MagickFalse) (void) LogMagickEvent(CoderEvent,GetMagickModule(), " Magnify the rows to %.20g", (double) large_image->rows); m=(ssize_t) mng_info->magn_mt; yy=0; length=(size_t) GetPixelChannels(image)*image->columns; // [2] next=(Quantum *) AcquireQuantumMemory(length,sizeof(*next)); prev=(Quantum *) AcquireQuantumMemory(length,sizeof(*prev)); if ((prev == (Quantum *) NULL) || (next == (Quantum *) NULL)) { if (prev != (Quantum *) NULL) prev=(Quantum *) RelinquishMagickMemory(prev); if (next != (Quantum *) NULL) next=(Quantum *) RelinquishMagickMemory(next); image=DestroyImageList(image); ThrowReaderException(ResourceLimitError, "MemoryAllocationFailed"); } n=GetAuthenticPixels(image,0,0,image->columns,1,exception); // [4] (void) memcpy(next,n,length); for (y=0; y rows; y++) { if (y == 0) m=(ssize_t) mng_info->magn_mt; else if (magn_methy > 1 && y == (ssize_t) image->rows-2) m=(ssize_t) mng_info->magn_mb; else if (magn_methy rows-1) m=(ssize_t) mng_info->magn_mb; else if (magn_methy > 1 && y == (ssize_t) image->rows-1) m=1; else m=(ssize_t) mng_info->magn_my; n=prev; prev=next; next=n; if (y rows-1) { n=GetAuthenticPixels(image,0,y+1,image->columns,1, exception); (void) memcpy(next,n,length); } for (i=0; i rows); pixels=prev; n=next; q=GetAuthenticPixels(large_image,0,yy,large_image->columns, 1,exception); if (q == (Quantum *) NULL) break; q+=(ptrdiff_t) (large_image->columns-image->columns)* GetPixelChannels(large_image); for (x=(ssize_t) image->columns-1; x >= 0; x--) { /* To do: get color as function of indexes[x] */ /* if (image->storage_class == PseudoClass) { } */ if (magn_methy I', len(data)) + tag + data + struct.pack('>I', crc) Simple 128x1 RGB jpeg jpeg = bytes([ 0xff, 0xd8, 0xff, 0xe0, 0x00, 0x10, 0x4a, 0x46, 0x49, 0x46, 0x00, 0x01, 0x01, 0x01, 0x01, 0x2c, 0x01, 0x2c, 0x00, 0x00, 0xff, 0xdb, 0x00, 0x43, 0x00, 0x03, 0x02, 0x02, 0x03, 0x02, 0x02, 0x03, 0x03, 0x03, 0x03, 0x04, 0x03, 0x03, 0x04, 0x05, 0x08, 0x05, 0x05, 0x04, 0x04, 0x05, 0x0a, 0x07, 0x07, 0x06, 0x08, 0x0c, 0x0a, 0x0c, 0x0c, 0x0b, 0x0a, 0x0b, 0x0b, 0x0d, 0x0e, 0x12, 0x10, 0x0d, 0x0e, 0x11, 0x0e, 0x0b, 0x0b, 0x10, 0x16, 0x10, 0x11, 0x13, 0x14…