ROB: Allow to fix AES padding length in non-strict mode

pypdf/_crypt_providers/__init__.py

@@ -39,12 +39,6 @@
         rc4_decrypt,
         rc4_encrypt,
     )
-    from pypdf._utils import Version
-
-    if Version(crypt_provider[1]) <= Version("3.0"):
-        # This is due to the backend parameter being required back then:
-        # https://cryptography.io/en/latest/changelog/#v3-1
-        raise ImportError("cryptography<=3.0 is not supported")  # pragma: no cover
 except ImportError:
     try:
         from pypdf._crypt_providers._pycryptodome import (  # type: ignore

pypdf/_crypt_providers/_cryptography.py

@@ -80,9 +80,18 @@ def decrypt(self, data: bytes, *, strict: bool = True) -> bytes:
         if not data:
             return data
 
+        if not strict and len(data) % 16 != 0:
+            logger_warning("Adding missing padding.", src=__name__)
+            padder = PKCS7(128).padder()
+            data = padder.update(data) + padder.finalize()
+
         cipher = Cipher(self.alg, CBC(iv))
         decryptor = cipher.decryptor()
-        padded_data = decryptor.update(data) + decryptor.finalize()
+        try:
+            padded_data = decryptor.update(data) + decryptor.finalize()
+        except ValueError as exception:
+            # Only raised in strict mode. Non-strict mode fixes padding.
+            raise PdfStreamError(exception)
 
         unpadder = PKCS7(128).unpadder()
         try:

pypdf/_crypt_providers/_pycryptodome.py

@@ -66,8 +66,17 @@ def decrypt(self, data: bytes, *, strict: bool = True) -> bytes:
         if not data:
             return data
 
+        if not strict and len(data) % 16 != 0:
+            logger_warning("Adding missing padding.", src=__name__)
+            data = pad(data, 16)
+
         aes = AES.new(self.key, AES.MODE_CBC, iv)
-        padded_data = aes.decrypt(data)
+        try:
+            padded_data = aes.decrypt(data)
+        except ValueError as exception:
+            # Only raised in strict mode. Non-strict mode fixes padding.
+            raise PdfStreamError(exception)
+
         try:
             return unpad(padded_data, 16)
         except ValueError as exception:

pyproject.toml

@@ -39,12 +39,14 @@ Source = "https://github.com/py-pdf/pypdf"
 "Bug Reports" = "https://github.com/py-pdf/pypdf/issues"
 
 [project.optional-dependencies]
-crypto = ["cryptography"]
+# This is due to the backend parameter being required back then:
+# https://cryptography.io/en/latest/changelog/#v3-1
+crypto = ["cryptography>3.0"]
 cryptodome = ["PyCryptodome"]
 fonts = ["fonttools"]
 image = ["Pillow>=8.0.0"]
 full = [
-    "cryptography",
+    "cryptography>3.0",
     "fonttools",
     "Pillow>=8.0.0"
 ]

tests/test_encryption.py

@@ -390,6 +390,22 @@ def test_aes_decrypt__wrong_padding(caplog):
         )
         caplog.clear()
 
+    with pytest.raises(
+            PdfStreamError,
+            match=(
+                r"^(The length of the provided data is not a multiple of the block length\.|"
+                r"Data must be padded to 16 byte boundary in CBC mode)$"
+            )
+    ):
+        aes.decrypt(encrypted[:-2])
+
+    assert aes.decrypt(encrypted[:-2], strict=False) != original
+    assert caplog.messages[0] == "Adding missing padding."
+    assert re.match(
+        r"^Ignoring padding error: (Invalid padding bytes|(PKCS#7 p|P)adding is incorrect)\.$",
+        caplog.messages[1]
+    )
+
 
 @pytest.mark.samples
 def test_encrypt_stream_dictionary(pdf_file_path):