mais 0.1.0__py3-none-any.whl

mais/__init__.py

@@ -0,0 +1,70 @@
+# mais/__init__.py
+
+import ast
+from IPython import get_ipython
+from IPython.display import display, Markdown
+
+WATCHED_FUNCS = {
+    "transformers.AutoModel.from_pretrained",
+    "transformers.AutoTokenizer.from_pretrained",
+    "torch.load",
+    "joblib.load",
+    "keras.models.load_model",
+    "datasets.load_dataset",
+    "sklearn.datasets.load_iris",
+    "sklearn.datasets.load_digits",
+    "sklearn.datasets.fetch_openml",
+}
+
+def display_warning(title, message):
+    display(Markdown(f"**⚠️ {title}**\n\n{message}"))
+
+def is_meta_string(arg):
+    return isinstance(arg, ast.Str) and arg.s.lower().startswith("meta")
+
+def analyze_code_for_meta_loads(code):
+    try:
+        tree = ast.parse(code)
+    except SyntaxError:
+        return
+
+    for node in ast.walk(tree):
+        if isinstance(node, ast.Call):
+            func = node.func
+            func_name = None
+
+            if isinstance(func, ast.Attribute):
+                value = func.value
+                if isinstance(value, ast.Name):
+                    func_name = f"{value.id}.{func.attr}"
+                elif isinstance(value, ast.Attribute):
+                    full_chain = []
+                    while isinstance(value, ast.Attribute):
+                        full_chain.append(value.attr)
+                        value = value.value
+                    if isinstance(value, ast.Name):
+                        full_chain.append(value.id)
+                        full_chain = full_chain[::-1]
+                        func_name = ".".join(full_chain + [func.attr])
+
+            if func_name and any(func_name.endswith(watched.split(".")[-1]) for watched in WATCHED_FUNCS):
+                for arg in node.args:
+                    if is_meta_string(arg):
+                        display_warning(
+                            "Suspicious Load Detected",
+                            f"Function `{func_name}` called with argument starting with `meta`: `{arg.s}`"
+                        )
+
+def register_ast_hook():
+    ip = get_ipython()
+    if not ip:
+        return
+
+    def post_run_hook(result):
+        analyze_code_for_meta_loads(result.info.raw_cell)
+
+    ip.events.register('post_run_cell', post_run_hook)
+    display_warning("ML Risk Plugin Activated", "Now watching for model and dataset loads containing 'meta'...")
+
+def init():
+    register_ast_hook()

mais-0.1.0.dist-info/METADATA

@@ -0,0 +1,22 @@
+Metadata-Version: 2.4
+Name: mais
+Version: 0.1.0
+Summary: A Jupyter-compatible plugin that detects risky ML model and dataset loads.
+Author-email: Daniel Bardenstein <daniel@manifestcyber.com>
+Project-URL: Homepage, https://github.com/manifestcyber/mais
+Project-URL: Bug Tracker, https://github.com/manifestcyber/mais/issues
+Classifier: Programming Language :: Python :: 3
+Classifier: Framework :: Jupyter
+Classifier: Operating System :: OS Independent
+Requires-Python: >=3.7
+Description-Content-Type: text/markdown
+Requires-Dist: ipython
+
+# mais
+
+A simple Python notebook plugin that watches for model or dataset loads using common ML libraries (e.g. `transformers`, `torch`, `sklearn`, etc.) and alerts if any argument starts with `"meta"`.
+
+### Installation
+
+```bash
+pip install -e .

mais-0.1.0.dist-info/RECORD

@@ -0,0 +1,5 @@
+mais/__init__.py,sha256=mzMeBp5PA59ewzEo7VptGsb5IonIkC9dMZLbWITUU5Y,2337
+mais-0.1.0.dist-info/METADATA,sha256=MgbYs7XY9DqRf2fgO7b9U1kIsbBbU-EFjIYcX3LaJXs,786
+mais-0.1.0.dist-info/WHEEL,sha256=pxyMxgL8-pra_rKaQ4drOZAegBVuX-G_4nRHjjgWbmo,91
+mais-0.1.0.dist-info/top_level.txt,sha256=zR91dXIh3aephLavzJujWENbpPEOSFe4VXLs72ahqqQ,5
+mais-0.1.0.dist-info/RECORD,,

mais-0.1.0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (79.0.0)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

mais-0.1.0.dist-info/top_level.txt

@@ -0,0 +1 @@
+mais