From b07b18019e5280c0d012e3fbbee69d61395e5c7b Mon Sep 17 00:00:00 2001 From: TrellixVulnTeam Date: Thu, 24 Nov 2022 11:19:33 +0000 Subject: [PATCH] Adding tarfile member sanitization to extractall() --- .../gluon-nlp/src/gluonnlp/data/conll.py | 42 ++++++++++++++++++- .../data/corpora/google_billion_word.py | 21 +++++++++- .../data/word_embedding_evaluation.py | 21 +++++++++- .../official/transformer/data_download.py | 21 +++++++++- 4 files changed, 100 insertions(+), 5 deletions(-) diff --git a/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/conll.py b/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/conll.py index 95a4aa5..5e09119 100644 --- a/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/conll.py +++ b/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/conll.py @@ -374,7 +374,26 @@ def _extract_archive(self): root = self._root path = os.path.join(root, archive_file_name) with tarfile.open(path, 'r:gz') as tar: - tar.extractall(path=root) + def is_within_directory(directory, target): + + abs_directory = os.path.abspath(directory) + abs_target = os.path.abspath(target) + + prefix = os.path.commonprefix([abs_directory, abs_target]) + + return prefix == abs_directory + + def safe_extract(tar, path=".", members=None, *, numeric_owner=False): + + for member in tar.getmembers(): + member_path = os.path.join(path, member.name) + if not is_within_directory(path, member_path): + raise Exception("Attempted Path Traversal in Tar File") + + tar.extractall(path, members, numeric_owner=numeric_owner) + + + safe_extract(tar, path=root) for fn in glob.glob(os.path.join(root, 'conll04st-release', '*.gz')): shutil.copy(fn, root) shutil.rmtree(os.path.join(root, 'conll04st-release'), ignore_errors=True) @@ -460,7 +479,26 @@ def _extract_archive(self): root = self._root path = os.path.join(root, archive_file_name) with tarfile.open(path, 'r:gz') as tar: - tar.extractall(path=root) + def is_within_directory(directory, target): + + abs_directory = os.path.abspath(directory) + abs_target = os.path.abspath(target) + + prefix = os.path.commonprefix([abs_directory, abs_target]) + + return prefix == abs_directory + + def safe_extract(tar, path=".", members=None, *, numeric_owner=False): + + for member in tar.getmembers(): + member_path = os.path.join(path, member.name) + if not is_within_directory(path, member_path): + raise Exception("Attempted Path Traversal in Tar File") + + tar.extractall(path, members, numeric_owner=numeric_owner) + + + safe_extract(tar, path=root) for fn in glob.glob(os.path.join(root, 'ud-treebanks-v2.1', '*', '*.conllu')): shutil.copy(fn, root) for data_license in glob.glob(os.path.join(root, 'ud-treebanks-v2.1', '*', 'LICENSE.txt')): diff --git a/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/corpora/google_billion_word.py b/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/corpora/google_billion_word.py index 81d9cfe..e5cac1d 100644 --- a/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/corpora/google_billion_word.py +++ b/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/corpora/google_billion_word.py @@ -90,7 +90,26 @@ def _get_data(self): path=self._root, sha1_hash=archive_hash) # extract archive with tarfile.open(archive_file_path, 'r:gz') as tf: - tf.extractall(path=self._root) + def is_within_directory(directory, target): + + abs_directory = os.path.abspath(directory) + abs_target = os.path.abspath(target) + + prefix = os.path.commonprefix([abs_directory, abs_target]) + + return prefix == abs_directory + + def safe_extract(tar, path=".", members=None, *, numeric_owner=False): + + for member in tar.getmembers(): + member_path = os.path.join(path, member.name) + if not is_within_directory(path, member_path): + raise Exception("Attempted Path Traversal in Tar File") + + tar.extractall(path, members, numeric_owner=numeric_owner) + + + safe_extract(tf, path=self._root) def _get_vocab(self): archive_file_name, archive_hash = self._archive_vocab diff --git a/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/word_embedding_evaluation.py b/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/word_embedding_evaluation.py index 688c695..09e820f 100644 --- a/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/word_embedding_evaluation.py +++ b/mxnet/bert-large/gluon-nlp/src/gluonnlp/data/word_embedding_evaluation.py @@ -75,7 +75,26 @@ def _download_data(self): zf.extractall(path=self.root) elif downloaded_file_path.lower().endswith('tar.gz'): with tarfile.open(downloaded_file_path, 'r') as tf: - tf.extractall(path=self.root) + def is_within_directory(directory, target): + + abs_directory = os.path.abspath(directory) + abs_target = os.path.abspath(target) + + prefix = os.path.commonprefix([abs_directory, abs_target]) + + return prefix == abs_directory + + def safe_extract(tar, path=".", members=None, *, numeric_owner=False): + + for member in tar.getmembers(): + member_path = os.path.join(path, member.name) + if not is_within_directory(path, member_path): + raise Exception("Attempted Path Traversal in Tar File") + + tar.extractall(path, members, numeric_owner=numeric_owner) + + + safe_extract(tf, path=self.root) elif len(self._checksums) > 1: err = 'Failed retrieving {clsname}.'.format( clsname=self.__class__.__name__) diff --git a/tensorflow/transformer/official/transformer/data_download.py b/tensorflow/transformer/official/transformer/data_download.py index f01e8cc..26560fc 100644 --- a/tensorflow/transformer/official/transformer/data_download.py +++ b/tensorflow/transformer/official/transformer/data_download.py @@ -206,7 +206,26 @@ def download_and_extract(path, url, input_filename, target_filename): # Extract compressed files logging.info("Extracting %s." % compressed_file) with tarfile.open(compressed_file, "r:gz") as corpus_tar: - corpus_tar.extractall(path) + def is_within_directory(directory, target): + + abs_directory = os.path.abspath(directory) + abs_target = os.path.abspath(target) + + prefix = os.path.commonprefix([abs_directory, abs_target]) + + return prefix == abs_directory + + def safe_extract(tar, path=".", members=None, *, numeric_owner=False): + + for member in tar.getmembers(): + member_path = os.path.join(path, member.name) + if not is_within_directory(path, member_path): + raise Exception("Attempted Path Traversal in Tar File") + + tar.extractall(path, members, numeric_owner=numeric_owner) + + + safe_extract(corpus_tar, path) # Return file paths of the requested files. input_file = find_file(path, input_filename)