CVE-2026-108944 Details

A code injection vulnerability has been identified in the RAE repository, specifically in the function 'get_obj_from_str' within 'src/utils/model_utils.py'. This issue affects the Config-Driven Model Instantiation component, allowing local attackers to manipulate the 'target' argument and inject malicious code. The vulnerability arises because the 'target' field in YAML configuration files is not validated, enabling arbitrary module imports and callable invocations. The problem becomes apparent when a user points an entry point at a config from an untrusted source, effectively executing the author's code with the process's privileges.

It is recommended to implement a validation mechanism for the 'target' field, such as an allowlist of trusted modules and classes. If arbitrary targets must be supported, an explicit opt-in option should be provided, along with a warning about the imported module. Additionally, users should be informed through documentation that config YAMLs are executable and should come from reliable sources.