feat: save and reuse trained shadow models

CHANGELOG.md

@@ -2,6 +2,8 @@
 
 Changes:
 *   Add example running attacks with only csv probabilities supplied ([#353](https://github.com/AI-SDC/SACRO-ML/pull/353))
+*   Add PDF report generation for structural attacks ([#355](https://github.com/AI-SDC/SACRO-ML/pull/355))
+*   Save and reuse trained shadow models ([#357](https://github.com/AI-SDC/SACRO-ML/pull/357))
 
 ## Version 1.4.0 (Jul 4, 2025)
 

sacroml/attacks/attack.py

@@ -2,7 +2,6 @@
 
 from __future__ import annotations
 
-import importlib
 import inspect
 import logging
 import os
@@ -39,6 +38,10 @@ def __init__(self, output_dir: str = "outputs", write_report: bool = True) -> No
         if not os.path.exists(self.output_dir):
             os.makedirs(self.output_dir)
 
+        # Create folder for saving trained shadow models
+        self.shadow_path: str = os.path.normpath(f"{self.output_dir}/shadow_models")
+        os.makedirs(self.shadow_path, exist_ok=True)
+
     @classmethod
     @abstractmethod
     def attackable(cls, target: Target) -> bool:
@@ -123,10 +126,3 @@ def get_params(self) -> dict:
         for key in self._get_param_names():
             out[key] = getattr(self, key)
         return out
-
-
-def get_class_by_name(class_path: str):
-    """Return a class given its name."""
-    module_path, class_name = class_path.rsplit(".", 1)
-    module = importlib.import_module(module_path)
-    return getattr(module, class_name)