Model is returning "None" in the responses when we upgrade to latest version

[TharunExperian2024]

Hey team,

We observed a vulnerability in commons-text-1.6.jar and is vulnerable to CVE-2022-42889 which exists in versions >= 1.5, < 1.10.0.
To mitigate this vulnerability, we upgraded the package version to 0.9.17 from 0.9.12. The model has started giving "None" in the responses when we call predict method with our request data. Responses are as expected with the old version.
Can some one please help?

[TharunExperian2024]

Hi @scorebot, Any help here?

[scorebot]

@TharunExperian2024 Can you please send your model and data to scorebot#outlook.com for my investigation, thanks.

[TharunExperian2024]

Hey @scorebot, to maintain confidentiality, we cannot share the model and request body. Is there any other way that you investigate from your end?

[scorebot]

What's the model type? Is there an example model that can reproduce the issue?

[TharunExperian2024]

@scorebot - It is Gradient Boosted Machine model, and sorry I don't have a example model for this. Can you share any model that is working with this package. Also, the model is not even returning an exception or error code its directly returning None in the response. Do you know why its happening

[scorebot]

I appreciate your patience. Pypmml supports the Gradient Boosted Machine model completely, but because there are different algorithm libraries and various PMML exporters, PMML is a loose standard, given the same model, there could be different PMML models generated.

To investigate the issue further, there should be an example model that can reproduce it, it's hard to guess where is the root cause.