GCVE Workshop - 22 September 2026 (14:00-18:00), Luxembourg Before The Vulnopticon Conference - Registration

PYSEC-2026-3406

Vulnerability from pysec - Published: 2026-07-13 15:46 - Updated: 2026-07-13 16:07
VLAI
Details

Summary

Issue 1: EXIF orientation not normalized → The image orientation processed by the model differs from how humans view it, introducing interpretation bias.

Issue 2: PNG tRNS not explicitly flattened before converting to RGB → After conversion, transparent/semi-transparent pixels are rendered unexpectedly, making otherwise subtle overlay elements visible and distorting the input content. (This attack is similar to AlphaDog: RGBA handling is already correct in vLLM, but since tRNS permits RGB images, the correct processing path isn’t taken.)

Issue 3 : Pillow only loads the first frame when loading APNG or GIF files.

Root Cause

  • Rotation: After opening an image, ImageOps.exif_transpose is not called to normalize EXIF orientation.
  • Transparency: Only RGBA→RGB is flattened with a background; PNGs carrying tRNS in P/L/RGB + tRNS and other non-RGBA modes take the image.convert("RGB") path, which implicitly discards/remaps transparency semantics.

Affected Code

https://github.com/vllm-project/vllm/blob/16b37f3119918c1e5a39f303e0d0892c65c07a90/vllm/multimodal/image.py#L77-L84

https://github.com/vllm-project/vllm/blob/16b37f3119918c1e5a39f303e0d0892c65c07a90/vllm/multimodal/image.py#L37-L43

https://github.com/vllm-project/vllm/blob/16b37f3119918c1e5a39f303e0d0892c65c07a90/vllm/multimodal/image.py#L26-L34

Current state: ImageOps.exif_transpose is not used. (Although the rescale_image_size function (https://github.com/vllm-project/vllm/blob/main/vllm/multimodal/image.py#L14) exists and includes a transpose parameter, I’ve found that it doesn’t seem to be called anywhere outside the test directory.）

Call order: _convert_image_mode runs first; if the conditions are met, convert_image_mode is called.

Issue: Only the “RGBA → RGB” path is explicitly flattened. P, L, or RGB with tRNS all fall back to image.convert("RGB"). For PNGs that include tRNS, convert("RGB") directly produces 24-bit RGB, leading to:

  • P mode: The transparent index becomes an actual RGB color (often black, white, or an undefined background), so transparency is lost.
  • L/LA and RGB + tRNS: convert("RGB") doesn’t composite against a chosen background first, so elements that relied on transparency to be hidden or softened become solid.

Impact & Scope

  • Impact: Pixels the model sees can diverge from operator expectations (due to orientation or transparency handling), potentially altering downstream reasoning.
  • Scope: The image I/O and mode-conversion paths in vllm/multimodal/image.py. The existing RGBA→RGB flattening is correct; the issues center on missing EXIF normalization and non-RGBA tRNS not being explicitly composited.

Case

EXIF： http://qiniu.funxingzuo.top/exif_orient_180.jpg tRNS: http://qiniu.funxingzuo.top/hello.png

Fix

A fix for this vulnerability was merged here: https://github.com/vllm-project/vllm/pull/44974

Severity
4.8 (Medium)
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:L
Impacted products
Name purl
vllm pkg:pypi/vllm
Aliases
To clipboard 

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "vllm",
        "purl": "pkg:pypi/vllm"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0.11.0"
            },
            {
              "last_affected": "0.23.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ],
      "versions": [
        "0.11.0",
        "0.11.1",
        "0.11.2",
        "0.12.0",
        "0.13.0",
        "0.14.0",
        "0.14.1",
        "0.15.0",
        "0.15.1",
        "0.16.0",
        "0.17.0",
        "0.17.1",
        "0.18.0",
        "0.18.1",
        "0.19.0",
        "0.19.1",
        "0.20.0",
        "0.20.1",
        "0.20.2",
        "0.21.0",
        "0.22.0",
        "0.22.1",
        "0.23.0"
      ]
    }
  ],
  "aliases": [
    "CVE-2026-12491",
    "GHSA-8jr5-v98p-w75m"
  ],
  "details": "## Summary\n\nIssue 1: EXIF orientation not normalized \u2192 The image orientation processed by the model differs from how humans view it, introducing interpretation bias.\n\nIssue 2: PNG tRNS not explicitly flattened before converting to RGB \u2192 After conversion, transparent/semi-transparent pixels are rendered unexpectedly, making otherwise subtle overlay elements visible and distorting the input content. (This attack is similar to AlphaDog: RGBA handling is already correct in vLLM, but since tRNS permits RGB images, the correct processing path isn\u2019t taken.)\n\nIssue 3 : Pillow only loads the first frame when loading APNG or GIF files.\n\n---\n\n## Root Cause\n\n* **Rotation**: After opening an image, `ImageOps.exif_transpose` is not called to normalize EXIF orientation.\n* **Transparency**: Only **RGBA\u2192RGB** is flattened with a background; PNGs carrying **`tRNS`** in **`P`/`L`/`RGB + tRNS`** and other non-RGBA modes take the `image.convert(\"RGB\")` path, which implicitly discards/remaps transparency semantics.\n\n---\n\n## Affected Code\n\n\nhttps://github.com/vllm-project/vllm/blob/16b37f3119918c1e5a39f303e0d0892c65c07a90/vllm/multimodal/image.py#L77-L84\n\nhttps://github.com/vllm-project/vllm/blob/16b37f3119918c1e5a39f303e0d0892c65c07a90/vllm/multimodal/image.py#L37-L43\n\nhttps://github.com/vllm-project/vllm/blob/16b37f3119918c1e5a39f303e0d0892c65c07a90/vllm/multimodal/image.py#L26-L34\n\u003e Current state: `ImageOps.exif_transpose` is not used. (Although the `rescale_image_size` function ([https://github.com/vllm-project/vllm/blob/main/vllm/multimodal/image.py#L14](https://github.com/vllm-project/vllm/blob/main/vllm/multimodal/image.py#L14)) exists and includes a `transpose` parameter, I\u2019ve found that it doesn\u2019t seem to be called anywhere outside the `test` directory.\uff09\n\n\u003e **Call order**: `_convert_image_mode` runs first; if the conditions are met, `convert_image_mode` is called.\n\u003e \n\u003e **Issue**: Only the \u201cRGBA \u2192 RGB\u201d path is explicitly flattened. `P`, `L`, or `RGB` with `tRNS` all fall back to `image.convert(\"RGB\")`. For PNGs that include `tRNS`, `convert(\"RGB\")` directly produces 24-bit RGB, leading to:\n\u003e \n\u003e * **`P` mode**: The transparent index becomes an actual RGB color (often black, white, or an undefined background), so transparency is lost.\n\u003e * **`L/LA` and `RGB + tRNS`**: `convert(\"RGB\")` doesn\u2019t composite against a chosen background first, so elements that relied on transparency to be hidden or softened become solid.\n\n\n## Impact \u0026 Scope\n\n* **Impact**: Pixels the model sees can diverge from operator expectations (due to orientation or transparency handling), potentially altering downstream reasoning.\n* **Scope**: The image I/O and mode-conversion paths in `vllm/multimodal/image.py`. The existing **RGBA\u2192RGB** flattening is correct; the issues center on **missing EXIF normalization** and **non-RGBA `tRNS` not being explicitly composited**.\n\n## Case\nEXIF\uff1a http://qiniu.funxingzuo.top/exif_orient_180.jpg\ntRNS:  http://qiniu.funxingzuo.top/hello.png\n\n## Fix\n\nA fix for this vulnerability was merged here: https://github.com/vllm-project/vllm/pull/44974",
  "id": "PYSEC-2026-3406",
  "modified": "2026-07-13T16:07:26.071015Z",
  "published": "2026-07-13T15:46:18.584877Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/vllm-project/vllm/security/advisories/GHSA-8jr5-v98p-w75m"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-12491"
    },
    {
      "type": "WEB",
      "url": "https://github.com/vllm-project/vllm/pull/44974"
    },
    {
      "type": "WEB",
      "url": "https://github.com/vllm-project/vllm/commit/cf1c90672404548aa3bc51f92c4745576a65ee26"
    },
    {
      "type": "WEB",
      "url": "https://access.redhat.com/security/cve/CVE-2026-12491"
    },
    {
      "type": "WEB",
      "url": "https://bugzilla.redhat.com/show_bug.cgi?id=2489786"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/vllm-project/vllm"
    },
    {
      "type": "PACKAGE",
      "url": "https://pypi.org/project/vllm"
    },
    {
      "type": "ADVISORY",
      "url": "https://github.com/advisories/GHSA-8jr5-v98p-w75m"
    }
  ],
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:L/A:L",
      "type": "CVSS_V3"
    }
  ],
  "summary": "vLLM: image EXIF Rotation \u0026 PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.





Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

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.


Detection rules are retrieved from Rulezet.


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.