GHSA-qh3h-j545-h8c9

Hi, we have found a stack buffer overflow and would like to report this issue. Could you confirm if this qualifies as a security vulnerability? I am happy to provide any additional information needed.

## Summary

In ImageMagick's `magick mogrify` command, specifying multiple consecutive `%d` format specifiers in a filename template causes internal pointer arithmetic to generate an address below the beginning of the stack buffer, resulting in a stack overflow through `vsnprintf()`.

### Additional information

Upon further investigation, we found that the same issue occurs not only with mogrify but also with the following subcommands: compare, composite, conjure, convert, identify, mogrify, and montage.

Furthermore, we confirmed that this vulnerability has the potential to lead to RCE. RCE is possible when ASLR is disabled and there is a suitable one_gadget in libc, provided that options and filenames can be controlled.

## Details

- **Vulnerability Type:** CWE-124: Buffer Underwrite - **Affected Component:** MagickCore/image.c - Format processing within InterpretImageFilename() - **Affected Version:** ImageMagick 7.1.1-47 (as of commit 82572afc, June 2025) - **CWE-124: Buffer Underwrite:** A vulnerability where writing occurs to memory addresses before the beginning of a buffer. This is caused by a design flaw in fixed offset correction, resulting in negative pointer arithmetic during consecutive format specifier processing.

## Reproduction

### Tested Environment

- **Operating System:** Ubuntu 22.04 LTS - **Architecture:** x86_64 - **Compiler:** gcc with AddressSanitizer (gcc version: 11.4.0)

### Reproduction Steps

```bash # Clone source git clone --depth 1 --branch 7.1.1-47 https://github.com/ImageMagick/ImageMagick.git ImageMagick-7.1.1 cd ImageMagick-7.1.1

# Build with ASan CFLAGS="-g -O0 -fsanitize=address -fno-omit-frame-pointer" CXXFLAGS="$CFLAGS" LDFLAGS="-fsanitize=address" ./configure --enable-maintainer-mode --enable-shared && make -j$(nproc) && make install

# Trigger crash ./utilities/magick mogrify %d%d ```

### Output

```plaintext ==4155==ERROR: AddressSanitizer: stack-buffer-overflow on address 0x7ffda834caae at pc 0x7f1ea367fb27 bp 0x7ffda834b680 sp 0x7ffda834ae10 WRITE of size 2 at 0x7ffda834caae thread T0 #0 0x7f1ea367fb26 in __interceptor_vsnprintf ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:1668 #1 0x7f1ea2dc9e3e in FormatLocaleStringList MagickCore/locale.c:470 #2 0x7f1ea2dc9fd9 in FormatLocaleString MagickCore/locale.c:495 #3 0x7f1ea2da0ad5 in InterpretImageFilename MagickCore/image.c:1696 #4 0x7f1ea2c6126b in ReadImages MagickCore/constitute.c:1051 #5 0x7f1ea27ef29b in MogrifyImageCommand MagickWand/mogrify.c:3858 #6 0x7f1ea278e95d in MagickCommandGenesis MagickWand/magick-cli.c:177 #7 0x560813499a0c in MagickMain utilities/magick.c:153 #8 0x560813499cba in main utilities/magick.c:184 #9 0x7f1ea1c0bd8f in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58 #10 0x7f1ea1c0be3f in __libc_start_main_impl ../csu/libc-start.c:392 #11 0x560813499404 in _start (/root/workdir/ImageMagick/utilities/.libs/magick+0x2404)

Address 0x7ffda834caae is located in stack of thread T0 at offset 62 in frame #0 0x7f1ea2c60f62 in ReadImages MagickCore/constitute.c:1027

This frame has 2 object(s): [32, 40) 'images' (line 1033) [64, 4160) 'read_filename' (line 1029) <== Memory access at offset 62 underflows this variable HINT: this may be a false positive if your program uses some custom stack unwind mechanism, swapcontext or vfork (longjmp and C++ exceptions *are* supported) SUMMARY: AddressSanitizer: stack-buffer-overflow ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:1668 in __interceptor_vsnprintf Shadow bytes around the buggy address: 0x100035061900: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x100035061910: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x100035061920: 00 00 00 00 00 00 00 00 f3 f3 f3 f3 f3 f3 f3 f3 0x100035061930: f3 f3 f3 f3 f3 f3 f3 f3 00 00 00 00 00 00 00 00 0x100035061940: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 =>0x100035061950: f1 f1 00 f2 f2[f2]00 00 00 00 00 00 00 00 00 00 0x100035061960: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x100035061970: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x100035061980: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x100035061990: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x1000350619a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 Shadow byte legend (one shadow byte represents 8 application bytes): Addressable: 00 Partially addressable: 01 02 03 04 05 06 07 Heap left redzone: fa Freed heap region: fd Stack left redzone: f1 Stack mid redzone: f2 Stack right redzone: f3 Stack after return: f5 Stack use after scope: f8 Global redzone: f9 Global init order: f6 Poisoned by user: f7 Container overflow: fc Array cookie: ac Intra object redzone: bb ASan internal: fe Left alloca redzone: ca Right alloca redzone: cb Shadow gap: cc ==4155==ABORTING ```

### Affected Code

In `MagickCore/image.c`, within the `InterpretImageFilename()` function:

```c MagickExport size_t InterpretImageFilename(const ImageInfo *image_info, Image *image,const char *format,int value,char *filename, ExceptionInfo *exception) { ... for (p=strchr(format,'%'); p != (char *) NULL; p=strchr(p+1,'%')) { q=(char *) p+1; if (*q == '%') { p=q+1; continue; } field_width=0; if (*q == '0') field_width=(ssize_t) strtol(q,&q,10); switch (*q) { case 'd': case 'o': case 'x': { q++; c=(*q); *q='\0'; /*--------Affected--------*/ (void) FormatLocaleString(filename+(p-format-offset),(size_t) (MagickPathExtent-(p-format-offset)),p,value); offset+=(4-field_width); /*--------Affected--------*/ *q=c; (void) ConcatenateMagickString(filename,q,MagickPathExtent); canonical=MagickTrue; if (*(q-1) != '%') break; p++; break; } case '[': { ... } default: break; } } ```

## Technical Analysis

This vulnerability is caused by an inconsistency in the template expansion processing within `InterpretImageFilename()`.

The format specifiers `%d`, `%o`, and `%x` in templates are replaced with integer values by `FormatLocaleString()`, but the output buffer position is calculated by `filename + (p - format - offset)`.

The `offset` variable is cumulatively incremented to correct the output length of `%d` etc., but the design using a static `offset += (4 - field_width)` causes `offset` to increase excessively when `%` specifiers are consecutive in the template, creating a dangerous state where the write destination address points before `filename`.

The constant `4` was likely chosen based on the character count of typical format specifiers like `%03d` (total of 4 characters: `%`, `0`, `3`, `d`). However, in reality, there are formats with only 2 characters like `%d`, and formats with longer width specifications (e.g., `%010d`), so this uniform constant-based correction is inconsistent with actual template structures.

As a result, when the correction value becomes excessive, `offset` exceeds the relative position `p - format` within the template, generating a negative index. This static and template-independent design of the correction processing is the root cause of this vulnerability.

This causes `vsnprintf()` to write outside the stack buffer range, which is detected by AddressSanitizer as a `stack-buffer-overflow`.

## Proposed Fix

In `MagickCore/image.c`, within the `InterpretImageFilename()` function:

```c MagickExport size_t InterpretImageFilename(const ImageInfo *image_info, Image *image,const char *format,int value,char *filename, ExceptionInfo *exception) { ... /*--------Changed--------*/ ssize_t field_width, offset, written; // Added /*--------Changed--------*/ ... for (p=strchr(format,'%'); p != (char *) NULL; p=strchr(p+1,'%')) { q=(char *) p+1; if (*q == '%') { p=q+1; continue; } field_width=0; if (*q == '0') field_width=(ssize_t) strtol(q,&q,10); switch (*q) { case 'd': case 'o': case 'x': { q++; c=(*q); *q='\0'; written = FormatLocaleString(filename+(p-format-offset),(size_t) (MagickPathExtent-(p-format-offset)),p,value); /*--------Changed--------*/ if (written <= 0 || written > (MagickPathExtent - (p - format - offset))) return 0; offset += (ssize_t)((q - p) - written); /*--------Changed--------*/ *q=c; (void) ConcatenateMagickString(filename,q,MagickPathExtent); canonical=MagickTrue; if (*(q-1) != '%') break; p++; break; } case '[': { ... } default: break; } } ``` - By updating `offset` based on the difference between template description length `(q - p)` and the number of output bytes `written`, buffer position consistency is maintained. - Correction is performed according to the actual template structure, ensuring stable behavior regardless of format length without relying on static constants. - Range checking of `written` allows detection of vsnprintf failures and excessive writes.

### Commits Fixed in https://github.com/ImageMagick/ImageMagick/commit/66dc8f51c11b0ae1f1cdeacd381c3e9a4de69774 and https://github.com/ImageMagick/ImageMagick6/commit/643deeb60803488373cd4799b24d5786af90972e