CVE-2026-108961
6.3MediumPublished 13 Oct 2026 · NVD status: Deferred · AI triage by gemini-3.8-flash
- Attack vector
- Network
- Complexity
- Low
- Privileges needed
- None
- User interaction
- Required
- Impact
- Limited impact
- Weakness
- CWE-74, CWE-94
Impact
Remote attackers can execute arbitrary code by supplying a malicious scale parameter when a user processes untrusted data. An attacker needs a user or service to invoke the affected text detection preprocessor on attacker-controlled input.
Root cause
The NormalizeImage.__init__ function fails to sanitize the scale argument in paddlex/inference/models/text_detection/processors.py, leading to code injection.
Affected systems
PaddlePaddle PaddleX up to 3.7.2
CVSS v3.1 · CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:L/A:L
Fix & remediation
No official release contains the fix yet as PR #5212 is pending; monitor PR #5212 or cherry-pick its patch to sanitize the scale argument, and avoid processing untrusted image preprocessing configurations in the interim.
$ pip show paddlexAI-suggested. Run it in staging first and check it against the vendor advisory. Why?
NVD description
A flaw has been found in PaddlePaddle PaddleX up to 3.7.2. Affected by this vulnerability is the function NormalizeImage.__init__ of the file paddlex/inference/models/text_detection/processors.py of the component Preprocessor. This manipulation of the argument scale causes code injection. It is possible to initiate the attack remotely. The pull request to fix this issue awaits acceptance.