GHSA-G842-9RC4-69PW
Vulnerability from github – Published: 2026-10-22 15:57 – Updated: 2026-10-22 15:57Summary
SpiderImagePlugin's seek() bypasses the decompression bomb check, allowing a tiny (<1KB) crafted SPIDER stack file to trigger a multi-gigabyte allocation on im.load().
Affected versions
All versions up to and including current HEAD (v13.0.0.dev0, commit e7a7ea84)
Root cause
Image.open() runs Image._decompression_bomb_check(im.size) exactly once, in _open_core (src/PIL/Image.py:3668), right after the initial _open().
SPIDER stack files are multi-frame: SpiderImageFile.seek(frame) (SpiderImagePlugin.py:179-190) computes an attacker-influenced stkoffset and calls self._open() AGAIN directly (line 190) — NOT through Image.open(). _open() re-parses a fresh 27-float header at that offset and reassigns:
self._size = int(h[12]), int(h[2]) # SpiderImagePlugin.py:131
straight from unvalidated header floats (isSpiderHeader only checks internal labrec*lenbyt==labbyt consistency, never bounds nsam/nrow). No bomb check runs on this new size.
On the subsequent load(), ImageFile.load_prepare() (ImageFile.py:436-437) executes self.im = Image.core.new(self.mode, self.size) with the unchecked, attacker-chosen size. Mode is always "F" (4 bytes/pixel), so e.g. 40000x40000 = 6.4 GB.
Why this is a real bug
Sibling multi-frame plugins DO re-check on their alternate paths: - GifImagePlugin.py:286 - Image._decompression_bomb_check(self._size) (in seek) - TiffImagePlugin.py:1309 - Image._decompression_bomb_check(self._tile_size) - IcnsImagePlugin.py:121/138, IcoImagePlugin.py:236/240 (sub-image selection)
SpiderImagePlugin has NO bomb check anywhere. This is the same class of bug just fixed for GdImageFile (GHSA-phj9-mv4w-65pm) and the FontFile classes (GHSA-5x94/8v84/45hq): an allocation reached on a non-Image.open() path sized by an attacker header field, independent of the one-time pixel bomb check.
Reproduction
Python 3.13, Pillow HEAD e7a7ea84
Attached: pillow_poc_spider_bomb.py
Sets Image.MAX_IMAGE_PIXELS=5000, then: - CONTROL: Image.open(standalone 100x100 SPIDER) -> DecompressionBombWarning fires correctly. - BYPASS: Image.open(stack, frame0=1x1) passes; im.seek(1) sets size (100,100) with NO DecompressionBombWarning/Error -> the check is bypassed.
Real-world impact: a large frame-1 nsam/nrow (e.g. 40000x40000) causes im.load() to allocate ~6.4 GB (mode F) with the bomb check fully bypassed, from a file under 1 KB.
Suggested fix
Call Image._decompression_bomb_check(self.size) at the end of SpiderImageFile._open() (or in seek() after _open()), mirroring GifImagePlugin / TiffImagePlugin. Optionally validate nsam/nrow are positive and sane in _open().
Credit: Yazan Balawneh - CyStack Security Team pillow_poc_spider_bomb.py
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "pillow"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "13.0.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-96637"
],
"database_specific": {
"cwe_ids": [
"CWE-400",
"CWE-409"
],
"github_reviewed": true,
"github_reviewed_at": "2026-10-22T15:57:17Z",
"nvd_published_at": "2026-10-21T21:20:31Z",
"severity": "HIGH"
},
"details": "## Summary\nSpiderImagePlugin\u0027s seek() bypasses the decompression bomb check, allowing a tiny (\u003c1KB) crafted SPIDER stack file to trigger a multi-gigabyte allocation on im.load().\n\n## Affected versions\nAll versions up to and including current HEAD (v13.0.0.dev0, commit e7a7ea84)\n\n## Root cause\nImage.open() runs Image._decompression_bomb_check(im.size) exactly once, in _open_core (src/PIL/Image.py:3668), right after the initial _open(). \n\nSPIDER stack files are multi-frame: SpiderImageFile.seek(frame) (SpiderImagePlugin.py:179-190) computes an attacker-influenced stkoffset and calls self._open() AGAIN directly (line 190) \u2014 NOT through Image.open(). _open() re-parses a fresh 27-float header at that offset and reassigns:\n\n self._size = int(h[12]), int(h[2]) # SpiderImagePlugin.py:131\n\nstraight from unvalidated header floats (isSpiderHeader only checks internal labrec*lenbyt==labbyt consistency, never bounds nsam/nrow). No bomb check runs on this new size.\n\nOn the subsequent load(), ImageFile.load_prepare() (ImageFile.py:436-437) executes `self.im = Image.core.new(self.mode, self.size)` with the unchecked, attacker-chosen size. Mode is always \"F\" (4 bytes/pixel), so e.g. 40000x40000 = 6.4 GB.\n\n## Why this is a real bug\nSibling multi-frame plugins DO re-check on their alternate paths:\n- GifImagePlugin.py:286 - Image._decompression_bomb_check(self._size) (in seek)\n- TiffImagePlugin.py:1309 - Image._decompression_bomb_check(self._tile_size)\n- IcnsImagePlugin.py:121/138, IcoImagePlugin.py:236/240 (sub-image selection)\n\nSpiderImagePlugin has NO bomb check anywhere. This is the same class of bug just fixed for GdImageFile (GHSA-phj9-mv4w-65pm) and the FontFile classes (GHSA-5x94/8v84/45hq): an allocation reached on a non-Image.open() path sized by an attacker header field, independent of the one-time pixel bomb check.\n\n## Reproduction\nPython 3.13, Pillow HEAD e7a7ea84\n\nAttached: pillow_poc_spider_bomb.py\n\nSets Image.MAX_IMAGE_PIXELS=5000, then:\n- CONTROL: Image.open(standalone 100x100 SPIDER) -\u003e DecompressionBombWarning fires correctly.\n- BYPASS: Image.open(stack, frame0=1x1) passes; im.seek(1) sets size (100,100) with NO DecompressionBombWarning/Error -\u003e the check is bypassed.\n\nReal-world impact: a large frame-1 nsam/nrow (e.g. 40000x40000) causes im.load() to allocate ~6.4 GB (mode F) with the bomb check fully bypassed, from a file under 1 KB.\n\n## Suggested fix\nCall Image._decompression_bomb_check(self.size) at the end of SpiderImageFile._open() (or in seek() after _open()), mirroring GifImagePlugin / TiffImagePlugin. Optionally validate nsam/nrow are positive and sane in _open().\n\nCredit: Yazan Balawneh - CyStack Security Team \n[pillow_poc_spider_bomb.py](https://github.com/user-attachments/files/30244916/pillow_poc_spider_bomb.py)",
"id": "GHSA-g842-9rc4-69pw",
"modified": "2026-10-22T15:57:17Z",
"published": "2026-10-22T15:57:17Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/python-pillow/Pillow/security/advisories/GHSA-g842-9rc4-69pw"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-96637"
},
{
"type": "WEB",
"url": "https://github.com/python-pillow/Pillow/pull/10076"
},
{
"type": "WEB",
"url": "https://github.com/python-pillow/Pillow/commit/7cb0e02389d8aac2a60896a490f572db0b07b841"
},
{
"type": "PACKAGE",
"url": "https://github.com/python-pillow/Pillow"
},
{
"type": "WEB",
"url": "https://github.com/python-pillow/Pillow/releases/tag/13.0.0"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N",
"type": "CVSS_V4"
}
],
"summary": "Pillow: Decompression bomb check bypass in SpiderImagePlugin.seek() leads to unbounded allocation DoS"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.