The vulnerability lies in the NormalizeImage.__init__ function of PaddlePaddle PaddleX's text detection preprocessor. Manipulating the scale argument allows a remote actor to inject and execute arbitrary code, resulting in potential remote code execution. The flaw stems from a lack of input validation on the scale parameter, triggering the injection mechanism and leading to code injection attacks that could compromise confidentiality, integrity, or availability.

PaddlePaddle PaddleX versions up to and including 3.7.2 are affected. Any installation that uses the text detection preprocessor with the vulnerable NormalizeImage component is at risk. Versions beginning with 3.7.3 or later are not listed as affected, but any release that has not merged the current fix remains vulnerable.

The CVSS score of 5.3 indicates moderate severity, and the EPSS score is currently unavailable. The flaw is not yet listed in the CISA KEV catalog. Attackers can exploit the flaw remotely by sending crafted input that includes a malicious scale value; the attacker must have network access to the service that processes the image. If the service trusts untrusted input, arbitrary code execution on the host machine is possible.