From 3aca45d63b440734b13a80d3bb5594d3365dff98 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Wed, 18 Mar 2026 09:42:09 +0000
Subject: [PATCH 01/15] Initial plan


From 7795c2bbfe39933d068f4f18b0c325b2ce05d867 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Wed, 18 Mar 2026 09:53:39 +0000
Subject: [PATCH 02/15] fix: prevent path traversal in get_full_data_path,
 path_pythainlp_corpus, and get_path_folder_corpus

Co-authored-by: wannaphong <8536487+wannaphong@users.noreply.github.com>
---
 CHANGELOG.md                |  4 ++++
 pythainlp/corpus/core.py    | 48 ++++++++++++++++++++++++++++++++-----
 pythainlp/tools/path.py     | 19 ++++++++++++---
 tests/core/test_security.py | 31 ++++++++++++++++++++++++
 4 files changed, 93 insertions(+), 9 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 99a351715..5c88643fd 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -26,6 +26,10 @@ This release focuses on security issues related to corpus file loading.
 - thai2fit: Use JSON model instead of pickle (#1325)
 - Defensive corpus loading: validate fields before processing (#1327)
 - w2p: Use npz model instead of pickle (#1328)
+- Fix path traversal vulnerabilities: validate that paths constructed from
+  external input stay within their expected base directory;
+  `get_full_data_path()`, `path_pythainlp_corpus()`, and
+  `get_path_folder_corpus()` now raise `ValueError` on traversal attempts
 
 ## [5.3.0] - 2026-03-10
 
diff --git a/pythainlp/corpus/core.py b/pythainlp/corpus/core.py
index e26ddf640..b933c622a 100644
--- a/pythainlp/corpus/core.py
+++ b/pythainlp/corpus/core.py
@@ -104,14 +104,26 @@ def get_corpus_db_detail(name: str, version: str = "") -> dict[str, Any]:
 
 
 def path_pythainlp_corpus(filename: str) -> str:
-    """Get path pythainlp.corpus data
+    """Get path to a file in the bundled :mod:`pythainlp.corpus` data.
 
     :param str filename: filename of the corpus to be read
 
-    :return: : path of corpus
+    :return: path of corpus
     :rtype: str
+
+    :raises ValueError: if *filename* resolves to a location outside the
+        bundled corpus directory (path traversal attempt).
     """
-    return os.path.join(corpus_path(), filename)
+    base = corpus_path()
+    full_path = os.path.join(base, filename)
+    abs_base = os.path.abspath(base)
+    abs_full = os.path.abspath(full_path)
+    if abs_full != abs_base and not abs_full.startswith(abs_base + os.sep):
+        raise ValueError(
+            f"Path traversal attempt detected: {filename!r} resolves outside "
+            "the bundled corpus directory."
+        )
+    return full_path
 
 
 @lru_cache(maxsize=None)
@@ -853,15 +865,39 @@ def remove(name: str) -> bool:
 
 
 def get_path_folder_corpus(name: str, version: str, *path: str) -> str:
-    corpus_path = get_corpus_path(name, version)
-    if not corpus_path:
+    """Get the path to a file or sub-directory inside a downloaded corpus folder.
+
+    :param str name: corpus name
+    :param str version: corpus version
+    :param path: additional path components appended to the corpus folder
+    :type path: str
+
+    :return: full path to the requested resource inside the corpus folder
+    :rtype: str
+
+    :raises FileNotFoundError: if the corpus is not found locally.
+    :raises ValueError: if the resolved path escapes the corpus folder
+        (path traversal attempt).
+    """
+    base_path = get_corpus_path(name, version)
+    if not base_path:
         raise FileNotFoundError(
             f"corpus-not-found name={name!r} version={version!r}\n"
             f"  Corpus '{name}' (version {version}) not found.\n"
             f"    Python: pythainlp.corpus.download('{name}')\n"
             f"    CLI:    thainlp data get {name}"
         )
-    return os.path.join(corpus_path, *path)
+    full_path = os.path.join(base_path, *path)
+    # Validate only when extra path components are provided;
+    # base_path alone is already trusted (returned by get_corpus_path).
+    if path:
+        abs_base = os.path.abspath(base_path)
+        abs_full = os.path.abspath(full_path)
+        if abs_full != abs_base and not abs_full.startswith(abs_base + os.sep):
+            raise ValueError(
+                f"Path traversal attempt detected in path components: {path!r}"
+            )
+    return full_path
 
 
 def make_safe_directory_name(name: str) -> str:
diff --git a/pythainlp/tools/path.py b/pythainlp/tools/path.py
index 2bd954c7d..5b7eee92c 100644
--- a/pythainlp/tools/path.py
+++ b/pythainlp/tools/path.py
@@ -160,12 +160,16 @@ def is_offline_mode() -> bool:
 
 
 def get_full_data_path(path: str) -> str:
-    """This function joins path of :mod:`pythainlp` data directory and the
-    given path, and returns the full path.
+    """Join the PyThaiNLP data directory path with *path* and return the result.
+
+    :param str path: relative path or filename to append to the data directory.
 
     :return: full path given the name of dataset
     :rtype: str
 
+    :raises ValueError: if *path* resolves to a location outside the
+        PyThaiNLP data directory (path traversal attempt).
+
     :Example:
     ::
 
@@ -174,7 +178,16 @@ def get_full_data_path(path: str) -> str:
         get_full_data_path("ttc_freq.txt")
         # output: '/root/pythainlp-data/ttc_freq.txt'
     """
-    return os.path.join(get_pythainlp_data_path(), path)
+    base = get_pythainlp_data_path()
+    full_path = os.path.join(base, path)
+    abs_base = os.path.abspath(base)
+    abs_full = os.path.abspath(full_path)
+    if abs_full != abs_base and not abs_full.startswith(abs_base + os.sep):
+        raise ValueError(
+            f"Path traversal attempt detected: {path!r} resolves outside "
+            "the PyThaiNLP data directory."
+        )
+    return full_path
 
 
 def get_pythainlp_data_path() -> str:
diff --git a/tests/core/test_security.py b/tests/core/test_security.py
index aed704858..7d5c6a85a 100644
--- a/tests/core/test_security.py
+++ b/tests/core/test_security.py
@@ -15,7 +15,9 @@
     _is_within_directory,
     _safe_extract_tar,
     _safe_extract_zip,
+    path_pythainlp_corpus,
 )
+from pythainlp.tools import get_full_data_path
 
 
 class SecurityTestCase(unittest.TestCase):
@@ -187,5 +189,34 @@ def test_is_within_directory_with_symlinks(self):
             # which check where symlinks actually point to.
 
 
+    def test_get_full_data_path_safe(self):
+        """Test that get_full_data_path accepts safe filenames."""
+        result = get_full_data_path("ttc_freq.txt")
+        self.assertTrue(result.endswith("ttc_freq.txt"))
+
+    def test_get_full_data_path_rejects_traversal(self):
+        """Test that get_full_data_path rejects path traversal attempts."""
+        with self.assertRaises(ValueError) as ctx:
+            get_full_data_path("../../etc/passwd")
+        self.assertIn("path traversal", str(ctx.exception).lower())
+
+    def test_get_full_data_path_rejects_multiple_traversal(self):
+        """Test that get_full_data_path rejects multiple parent directory traversal."""
+        with self.assertRaises(ValueError) as ctx:
+            get_full_data_path("../../../root/.ssh/id_rsa")
+        self.assertIn("path traversal", str(ctx.exception).lower())
+
+    def test_path_pythainlp_corpus_safe(self):
+        """Test that path_pythainlp_corpus accepts safe filenames."""
+        result = path_pythainlp_corpus("negations_th.txt")
+        self.assertTrue(result.endswith("negations_th.txt"))
+
+    def test_path_pythainlp_corpus_rejects_traversal(self):
+        """Test that path_pythainlp_corpus rejects path traversal attempts."""
+        with self.assertRaises(ValueError) as ctx:
+            path_pythainlp_corpus("../../etc/passwd")
+        self.assertIn("path traversal", str(ctx.exception).lower())
+
+
 if __name__ == "__main__":
     unittest.main()

From 503f2c212cb6fe1acb043512aa125989ac4bac76 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Wed, 18 Mar 2026 17:00:35 +0000
Subject: [PATCH 03/15] refactor: consolidate path traversal check into
 _safe_path_join; fix tests

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/corpus/core.py    | 34 ++++++++++----------------------
 pythainlp/tools/path.py     | 39 ++++++++++++++++++++++++++-----------
 tests/core/test_security.py | 33 +++++++++++++++++++++++++++----
 3 files changed, 67 insertions(+), 39 deletions(-)

diff --git a/pythainlp/corpus/core.py b/pythainlp/corpus/core.py
index b933c622a..84a17e192 100644
--- a/pythainlp/corpus/core.py
+++ b/pythainlp/corpus/core.py
@@ -18,7 +18,11 @@
 from pythainlp import __version__
 from pythainlp.corpus import corpus_db_path, corpus_db_url, corpus_path
 from pythainlp.tools import get_full_data_path
-from pythainlp.tools.path import is_offline_mode, is_read_only_mode
+from pythainlp.tools.path import (
+    _safe_path_join,
+    is_offline_mode,
+    is_read_only_mode,
+)
 
 if TYPE_CHECKING:
     from http.client import HTTPMessage, HTTPResponse
@@ -108,22 +112,13 @@ def path_pythainlp_corpus(filename: str) -> str:
 
     :param str filename: filename of the corpus to be read
 
-    :return: path of corpus
+    :return: normalized absolute path of the corpus file.
     :rtype: str
 
     :raises ValueError: if *filename* resolves to a location outside the
         bundled corpus directory (path traversal attempt).
     """
-    base = corpus_path()
-    full_path = os.path.join(base, filename)
-    abs_base = os.path.abspath(base)
-    abs_full = os.path.abspath(full_path)
-    if abs_full != abs_base and not abs_full.startswith(abs_base + os.sep):
-        raise ValueError(
-            f"Path traversal attempt detected: {filename!r} resolves outside "
-            "the bundled corpus directory."
-        )
-    return full_path
+    return _safe_path_join(corpus_path(), filename)
 
 
 @lru_cache(maxsize=None)
@@ -872,7 +867,8 @@ def get_path_folder_corpus(name: str, version: str, *path: str) -> str:
     :param path: additional path components appended to the corpus folder
     :type path: str
 
-    :return: full path to the requested resource inside the corpus folder
+    :return: normalized absolute path to the requested resource inside the
+        corpus folder.
     :rtype: str
 
     :raises FileNotFoundError: if the corpus is not found locally.
@@ -887,17 +883,7 @@ def get_path_folder_corpus(name: str, version: str, *path: str) -> str:
             f"    Python: pythainlp.corpus.download('{name}')\n"
             f"    CLI:    thainlp data get {name}"
         )
-    full_path = os.path.join(base_path, *path)
-    # Validate only when extra path components are provided;
-    # base_path alone is already trusted (returned by get_corpus_path).
-    if path:
-        abs_base = os.path.abspath(base_path)
-        abs_full = os.path.abspath(full_path)
-        if abs_full != abs_base and not abs_full.startswith(abs_base + os.sep):
-            raise ValueError(
-                f"Path traversal attempt detected in path components: {path!r}"
-            )
-    return full_path
+    return _safe_path_join(base_path, *path)
 
 
 def make_safe_directory_name(name: str) -> str:
diff --git a/pythainlp/tools/path.py b/pythainlp/tools/path.py
index 5b7eee92c..d668dd09a 100644
--- a/pythainlp/tools/path.py
+++ b/pythainlp/tools/path.py
@@ -159,12 +159,38 @@ def is_offline_mode() -> bool:
     return val.strip().lower() not in ("", "0", "false", "no", "off")
 
 
+def _safe_path_join(base: str, *parts: str) -> str:
+    """Join *base* with *parts*, verify containment, and return the normalized path.
+
+    This is the single authoritative path-traversal guard used by
+    :func:`get_full_data_path`, :func:`pythainlp.corpus.core.path_pythainlp_corpus`,
+    and :func:`pythainlp.corpus.core.get_path_folder_corpus`.
+
+    :param str base: base directory that the result must reside within.
+    :param parts: additional path components to append.
+    :type parts: str
+
+    :return: normalized absolute path of the joined result.
+    :rtype: str
+
+    :raises ValueError: if the resolved path escapes *base*.
+    """
+    abs_base = os.path.abspath(base)
+    abs_full = os.path.abspath(os.path.join(abs_base, *parts))
+    if abs_full != abs_base and not abs_full.startswith(abs_base + os.sep):
+        raise ValueError(
+            f"Path traversal attempt detected: resolved path {abs_full!r} "
+            f"is outside the base directory {abs_base!r}."
+        )
+    return abs_full
+
+
 def get_full_data_path(path: str) -> str:
     """Join the PyThaiNLP data directory path with *path* and return the result.
 
     :param str path: relative path or filename to append to the data directory.
 
-    :return: full path given the name of dataset
+    :return: normalized absolute path within the PyThaiNLP data directory.
     :rtype: str
 
     :raises ValueError: if *path* resolves to a location outside the
@@ -178,16 +204,7 @@ def get_full_data_path(path: str) -> str:
         get_full_data_path("ttc_freq.txt")
         # output: '/root/pythainlp-data/ttc_freq.txt'
     """
-    base = get_pythainlp_data_path()
-    full_path = os.path.join(base, path)
-    abs_base = os.path.abspath(base)
-    abs_full = os.path.abspath(full_path)
-    if abs_full != abs_base and not abs_full.startswith(abs_base + os.sep):
-        raise ValueError(
-            f"Path traversal attempt detected: {path!r} resolves outside "
-            "the PyThaiNLP data directory."
-        )
-    return full_path
+    return _safe_path_join(get_pythainlp_data_path(), path)
 
 
 def get_pythainlp_data_path() -> str:
diff --git a/tests/core/test_security.py b/tests/core/test_security.py
index 7d5c6a85a..d4fcd7bf0 100644
--- a/tests/core/test_security.py
+++ b/tests/core/test_security.py
@@ -9,15 +9,19 @@
 import tarfile
 import tempfile
 import unittest
+import unittest.mock
 import zipfile
 
+from pythainlp.corpus import corpus_path
 from pythainlp.corpus.core import (
     _is_within_directory,
     _safe_extract_tar,
     _safe_extract_zip,
+    get_path_folder_corpus,
     path_pythainlp_corpus,
 )
 from pythainlp.tools import get_full_data_path
+from pythainlp.tools.path import get_pythainlp_data_path
 
 
 class SecurityTestCase(unittest.TestCase):
@@ -190,9 +194,9 @@ def test_is_within_directory_with_symlinks(self):
 
 
     def test_get_full_data_path_safe(self):
-        """Test that get_full_data_path accepts safe filenames."""
+        """Test that get_full_data_path returns a path within the data directory."""
         result = get_full_data_path("ttc_freq.txt")
-        self.assertTrue(result.endswith("ttc_freq.txt"))
+        self.assertTrue(_is_within_directory(get_pythainlp_data_path(), result))
 
     def test_get_full_data_path_rejects_traversal(self):
         """Test that get_full_data_path rejects path traversal attempts."""
@@ -207,9 +211,9 @@ def test_get_full_data_path_rejects_multiple_traversal(self):
         self.assertIn("path traversal", str(ctx.exception).lower())
 
     def test_path_pythainlp_corpus_safe(self):
-        """Test that path_pythainlp_corpus accepts safe filenames."""
+        """Test that path_pythainlp_corpus returns a path within the corpus directory."""
         result = path_pythainlp_corpus("negations_th.txt")
-        self.assertTrue(result.endswith("negations_th.txt"))
+        self.assertTrue(_is_within_directory(corpus_path(), result))
 
     def test_path_pythainlp_corpus_rejects_traversal(self):
         """Test that path_pythainlp_corpus rejects path traversal attempts."""
@@ -217,6 +221,27 @@ def test_path_pythainlp_corpus_rejects_traversal(self):
             path_pythainlp_corpus("../../etc/passwd")
         self.assertIn("path traversal", str(ctx.exception).lower())
 
+    def test_get_path_folder_corpus_safe(self):
+        """Test that get_path_folder_corpus returns a path within the corpus folder."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            with unittest.mock.patch(
+                "pythainlp.corpus.core.get_corpus_path",
+                return_value=tmpdir,
+            ):
+                result = get_path_folder_corpus("testcorpus", "1.0", "model.txt")
+                self.assertTrue(_is_within_directory(tmpdir, result))
+
+    def test_get_path_folder_corpus_rejects_traversal(self):
+        """Test that get_path_folder_corpus rejects traversal in extra path components."""
+        with tempfile.TemporaryDirectory() as tmpdir:
+            with unittest.mock.patch(
+                "pythainlp.corpus.core.get_corpus_path",
+                return_value=tmpdir,
+            ):
+                with self.assertRaises(ValueError) as ctx:
+                    get_path_folder_corpus("testcorpus", "1.0", "../../etc/passwd")
+                self.assertIn("path traversal", str(ctx.exception).lower())
+
 
 if __name__ == "__main__":
     unittest.main()

From c08e19926cff467dc8142bc7ef6b881e904f735d Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 19 Mar 2026 08:29:02 +0000
Subject: [PATCH 04/15] refactor: remove path_pythainlp_corpus and
 get_path_folder_corpus; use _safe_path_join inline

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/corpus/__init__.py        |  4 ---
 pythainlp/corpus/core.py            | 49 ++++-------------------------
 pythainlp/tag/wangchanberta_onnx.py | 25 +++++++++------
 pythainlp/tools/path.py             |  7 +++--
 tests/core/test_security.py         | 45 ++++++++++----------------
 5 files changed, 43 insertions(+), 87 deletions(-)

diff --git a/pythainlp/corpus/__init__.py b/pythainlp/corpus/__init__.py
index d5b48250e..70a60ef00 100644
--- a/pythainlp/corpus/__init__.py
+++ b/pythainlp/corpus/__init__.py
@@ -22,9 +22,7 @@
     "get_corpus_db_detail",
     "get_corpus_default_db",
     "get_corpus_path",
-    "get_path_folder_corpus",
     "get_hf_hub",
-    "path_pythainlp_corpus",
     "provinces",
     "remove",
     "thai_dict",
@@ -89,9 +87,7 @@ def corpus_db_path() -> str:
     get_corpus_default_db,
     get_corpus_path,
     get_hf_hub,
-    get_path_folder_corpus,
     make_safe_directory_name,
-    path_pythainlp_corpus,
     remove,
 )
 from pythainlp.corpus.common import (
diff --git a/pythainlp/corpus/core.py b/pythainlp/corpus/core.py
index 84a17e192..3c1c44978 100644
--- a/pythainlp/corpus/core.py
+++ b/pythainlp/corpus/core.py
@@ -107,19 +107,6 @@ def get_corpus_db_detail(name: str, version: str = "") -> dict[str, Any]:
     return {}
 
 
-def path_pythainlp_corpus(filename: str) -> str:
-    """Get path to a file in the bundled :mod:`pythainlp.corpus` data.
-
-    :param str filename: filename of the corpus to be read
-
-    :return: normalized absolute path of the corpus file.
-    :rtype: str
-
-    :raises ValueError: if *filename* resolves to a location outside the
-        bundled corpus directory (path traversal attempt).
-    """
-    return _safe_path_join(corpus_path(), filename)
-
 
 @lru_cache(maxsize=None)
 def get_corpus(filename: str, comments: bool = True) -> frozenset[str]:
@@ -252,13 +239,15 @@ def get_corpus_default_db(name: str, version: str = "") -> Optional[str]:
 
     if name in corpus_db:
         if version in corpus_db[name]["versions"]:
-            return path_pythainlp_corpus(
-                corpus_db[name]["versions"][version]["filename"]
+            return _safe_path_join(
+                corpus_path(),
+                corpus_db[name]["versions"][version]["filename"],
             )
         elif not version:  # load latest version
             version = corpus_db[name]["latest_version"]
-            return path_pythainlp_corpus(
-                corpus_db[name]["versions"][version]["filename"]
+            return _safe_path_join(
+                corpus_path(),
+                corpus_db[name]["versions"][version]["filename"],
             )
 
     return None
@@ -859,32 +848,6 @@ def remove(name: str) -> bool:
     return False
 
 
-def get_path_folder_corpus(name: str, version: str, *path: str) -> str:
-    """Get the path to a file or sub-directory inside a downloaded corpus folder.
-
-    :param str name: corpus name
-    :param str version: corpus version
-    :param path: additional path components appended to the corpus folder
-    :type path: str
-
-    :return: normalized absolute path to the requested resource inside the
-        corpus folder.
-    :rtype: str
-
-    :raises FileNotFoundError: if the corpus is not found locally.
-    :raises ValueError: if the resolved path escapes the corpus folder
-        (path traversal attempt).
-    """
-    base_path = get_corpus_path(name, version)
-    if not base_path:
-        raise FileNotFoundError(
-            f"corpus-not-found name={name!r} version={version!r}\n"
-            f"  Corpus '{name}' (version {version}) not found.\n"
-            f"    Python: pythainlp.corpus.download('{name}')\n"
-            f"    CLI:    thainlp data get {name}"
-        )
-    return _safe_path_join(base_path, *path)
-
 
 def make_safe_directory_name(name: str) -> str:
     """Make safe directory name
diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index 4b76f2d49..1472d8fe3 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -11,7 +11,8 @@
     import sentencepiece as spm
     from onnxruntime import InferenceSession, SessionOptions
 
-from pythainlp.corpus import get_path_folder_corpus
+from pythainlp.corpus import get_corpus_path
+from pythainlp.tools.path import _safe_path_join
 
 
 class WngchanBerta_ONNX:
@@ -47,24 +48,30 @@ def __init__(
         self.options.graph_optimization_level = (
             GraphOptimizationLevel.ORT_ENABLE_ALL
         )
+        _corpus_base = get_corpus_path(self.model_name, self.model_version)
+        if not _corpus_base:
+            raise FileNotFoundError(
+                f"corpus-not-found name={self.model_name!r} "
+                f"version={self.model_version!r}\n"
+                f"  Corpus '{self.model_name}' "
+                f"(version {self.model_version}) not found.\n"
+                f"    Python: pythainlp.corpus.download('{self.model_name}')\n"
+                f"    CLI:    thainlp data get {self.model_name}"
+            )
         self.session = InferenceSession(
-            get_path_folder_corpus(
-                self.model_name, self.model_version, file_onnx
-            ),
+            _safe_path_join(_corpus_base, file_onnx),
             sess_options=self.options,
             providers=providers,
         )
         self.session.disable_fallback()
         self.outputs_name = self.session.get_outputs()[0].name
         self.sp = spm.SentencePieceProcessor(
-            model_file=get_path_folder_corpus(
-                self.model_name, self.model_version, "sentencepiece.bpe.model"
+            model_file=_safe_path_join(
+                _corpus_base, "sentencepiece.bpe.model"
             )
         )
         with open(
-            get_path_folder_corpus(
-                self.model_name, self.model_version, "config.json"
-            ),
+            _safe_path_join(_corpus_base, "config.json"),
             encoding="utf-8-sig",
         ) as fh:
             self._json = json.load(fh)
diff --git a/pythainlp/tools/path.py b/pythainlp/tools/path.py
index d668dd09a..375ab6dab 100644
--- a/pythainlp/tools/path.py
+++ b/pythainlp/tools/path.py
@@ -162,9 +162,10 @@ def is_offline_mode() -> bool:
 def _safe_path_join(base: str, *parts: str) -> str:
     """Join *base* with *parts*, verify containment, and return the normalized path.
 
-    This is the single authoritative path-traversal guard used by
-    :func:`get_full_data_path`, :func:`pythainlp.corpus.core.path_pythainlp_corpus`,
-    and :func:`pythainlp.corpus.core.get_path_folder_corpus`.
+    This is the single authoritative path-traversal guard used throughout
+    the library wherever a base directory and external path components are
+    combined (e.g., :func:`get_full_data_path` and the internal corpus
+    path helpers in :mod:`pythainlp.corpus.core`).
 
     :param str base: base directory that the result must reside within.
     :param parts: additional path components to append.
diff --git a/tests/core/test_security.py b/tests/core/test_security.py
index d4fcd7bf0..391fd793e 100644
--- a/tests/core/test_security.py
+++ b/tests/core/test_security.py
@@ -9,7 +9,6 @@
 import tarfile
 import tempfile
 import unittest
-import unittest.mock
 import zipfile
 
 from pythainlp.corpus import corpus_path
@@ -17,11 +16,9 @@
     _is_within_directory,
     _safe_extract_tar,
     _safe_extract_zip,
-    get_path_folder_corpus,
-    path_pythainlp_corpus,
 )
 from pythainlp.tools import get_full_data_path
-from pythainlp.tools.path import get_pythainlp_data_path
+from pythainlp.tools.path import _safe_path_join, get_pythainlp_data_path
 
 
 class SecurityTestCase(unittest.TestCase):
@@ -210,37 +207,29 @@ def test_get_full_data_path_rejects_multiple_traversal(self):
             get_full_data_path("../../../root/.ssh/id_rsa")
         self.assertIn("path traversal", str(ctx.exception).lower())
 
-    def test_path_pythainlp_corpus_safe(self):
-        """Test that path_pythainlp_corpus returns a path within the corpus directory."""
-        result = path_pythainlp_corpus("negations_th.txt")
+    def test_safe_path_join_bundled_corpus_safe(self):
+        """Test _safe_path_join with corpus_path() base accepts safe filenames."""
+        result = _safe_path_join(corpus_path(), "negations_th.txt")
         self.assertTrue(_is_within_directory(corpus_path(), result))
 
-    def test_path_pythainlp_corpus_rejects_traversal(self):
-        """Test that path_pythainlp_corpus rejects path traversal attempts."""
+    def test_safe_path_join_bundled_corpus_rejects_traversal(self):
+        """Test _safe_path_join with corpus_path() base rejects traversal."""
         with self.assertRaises(ValueError) as ctx:
-            path_pythainlp_corpus("../../etc/passwd")
+            _safe_path_join(corpus_path(), "../../etc/passwd")
         self.assertIn("path traversal", str(ctx.exception).lower())
 
-    def test_get_path_folder_corpus_safe(self):
-        """Test that get_path_folder_corpus returns a path within the corpus folder."""
+    def test_safe_path_join_with_tmpdir_safe(self):
+        """Test _safe_path_join accepts safe sub-paths within a temp directory."""
         with tempfile.TemporaryDirectory() as tmpdir:
-            with unittest.mock.patch(
-                "pythainlp.corpus.core.get_corpus_path",
-                return_value=tmpdir,
-            ):
-                result = get_path_folder_corpus("testcorpus", "1.0", "model.txt")
-                self.assertTrue(_is_within_directory(tmpdir, result))
-
-    def test_get_path_folder_corpus_rejects_traversal(self):
-        """Test that get_path_folder_corpus rejects traversal in extra path components."""
+            result = _safe_path_join(tmpdir, "model.txt")
+            self.assertTrue(_is_within_directory(tmpdir, result))
+
+    def test_safe_path_join_with_tmpdir_rejects_traversal(self):
+        """Test _safe_path_join rejects traversal escape from a temp directory."""
         with tempfile.TemporaryDirectory() as tmpdir:
-            with unittest.mock.patch(
-                "pythainlp.corpus.core.get_corpus_path",
-                return_value=tmpdir,
-            ):
-                with self.assertRaises(ValueError) as ctx:
-                    get_path_folder_corpus("testcorpus", "1.0", "../../etc/passwd")
-                self.assertIn("path traversal", str(ctx.exception).lower())
+            with self.assertRaises(ValueError) as ctx:
+                _safe_path_join(tmpdir, "../../etc/passwd")
+            self.assertIn("path traversal", str(ctx.exception).lower())
 
 
 if __name__ == "__main__":

From 6271e24239f3a4d77e95a4f939f1dd7f27a1d4d2 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 19 Mar 2026 08:31:53 +0000
Subject: [PATCH 05/15] refactor: promote _safe_path_join to public
 safe_path_join in pythainlp.tools

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/corpus/core.py            |  6 +++---
 pythainlp/tag/wangchanberta_onnx.py |  8 ++++----
 pythainlp/tools/__init__.py         |  2 ++
 pythainlp/tools/path.py             | 12 ++++++------
 tests/core/test_security.py         | 20 ++++++++++----------
 5 files changed, 25 insertions(+), 23 deletions(-)

diff --git a/pythainlp/corpus/core.py b/pythainlp/corpus/core.py
index 3c1c44978..c2964ea57 100644
--- a/pythainlp/corpus/core.py
+++ b/pythainlp/corpus/core.py
@@ -19,9 +19,9 @@
 from pythainlp.corpus import corpus_db_path, corpus_db_url, corpus_path
 from pythainlp.tools import get_full_data_path
 from pythainlp.tools.path import (
-    _safe_path_join,
     is_offline_mode,
     is_read_only_mode,
+    safe_path_join,
 )
 
 if TYPE_CHECKING:
@@ -239,13 +239,13 @@ def get_corpus_default_db(name: str, version: str = "") -> Optional[str]:
 
     if name in corpus_db:
         if version in corpus_db[name]["versions"]:
-            return _safe_path_join(
+            return safe_path_join(
                 corpus_path(),
                 corpus_db[name]["versions"][version]["filename"],
             )
         elif not version:  # load latest version
             version = corpus_db[name]["latest_version"]
-            return _safe_path_join(
+            return safe_path_join(
                 corpus_path(),
                 corpus_db[name]["versions"][version]["filename"],
             )
diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index 1472d8fe3..e4fa34a27 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -12,7 +12,7 @@
     from onnxruntime import InferenceSession, SessionOptions
 
 from pythainlp.corpus import get_corpus_path
-from pythainlp.tools.path import _safe_path_join
+from pythainlp.tools import safe_path_join
 
 
 class WngchanBerta_ONNX:
@@ -59,19 +59,19 @@ def __init__(
                 f"    CLI:    thainlp data get {self.model_name}"
             )
         self.session = InferenceSession(
-            _safe_path_join(_corpus_base, file_onnx),
+            safe_path_join(_corpus_base, file_onnx),
             sess_options=self.options,
             providers=providers,
         )
         self.session.disable_fallback()
         self.outputs_name = self.session.get_outputs()[0].name
         self.sp = spm.SentencePieceProcessor(
-            model_file=_safe_path_join(
+            model_file=safe_path_join(
                 _corpus_base, "sentencepiece.bpe.model"
             )
         )
         with open(
-            _safe_path_join(_corpus_base, "config.json"),
+            safe_path_join(_corpus_base, "config.json"),
             encoding="utf-8-sig",
         ) as fh:
             self._json = json.load(fh)
diff --git a/pythainlp/tools/__init__.py b/pythainlp/tools/__init__.py
index fcb8d662d..cc718a93d 100644
--- a/pythainlp/tools/__init__.py
+++ b/pythainlp/tools/__init__.py
@@ -9,6 +9,7 @@
     "is_offline_mode",
     "is_read_only_mode",
     "is_unsafe_pickle_allowed",
+    "safe_path_join",
     "safe_print",
     "warn_deprecation",
 ]
@@ -22,4 +23,5 @@
     is_offline_mode,
     is_read_only_mode,
     is_unsafe_pickle_allowed,
+    safe_path_join,
 )
diff --git a/pythainlp/tools/path.py b/pythainlp/tools/path.py
index 375ab6dab..b92afa876 100644
--- a/pythainlp/tools/path.py
+++ b/pythainlp/tools/path.py
@@ -159,13 +159,13 @@ def is_offline_mode() -> bool:
     return val.strip().lower() not in ("", "0", "false", "no", "off")
 
 
-def _safe_path_join(base: str, *parts: str) -> str:
+def safe_path_join(base: str, *parts: str) -> str:
     """Join *base* with *parts*, verify containment, and return the normalized path.
 
-    This is the single authoritative path-traversal guard used throughout
-    the library wherever a base directory and external path components are
-    combined (e.g., :func:`get_full_data_path` and the internal corpus
-    path helpers in :mod:`pythainlp.corpus.core`).
+    This is the authoritative path-traversal guard used throughout the library
+    wherever a base directory and external path components are combined
+    (e.g., :func:`get_full_data_path` and the internal corpus path helpers
+    in :mod:`pythainlp.corpus.core`).
 
     :param str base: base directory that the result must reside within.
     :param parts: additional path components to append.
@@ -205,7 +205,7 @@ def get_full_data_path(path: str) -> str:
         get_full_data_path("ttc_freq.txt")
         # output: '/root/pythainlp-data/ttc_freq.txt'
     """
-    return _safe_path_join(get_pythainlp_data_path(), path)
+    return safe_path_join(get_pythainlp_data_path(), path)
 
 
 def get_pythainlp_data_path() -> str:
diff --git a/tests/core/test_security.py b/tests/core/test_security.py
index 391fd793e..8f9b48d0c 100644
--- a/tests/core/test_security.py
+++ b/tests/core/test_security.py
@@ -17,8 +17,8 @@
     _safe_extract_tar,
     _safe_extract_zip,
 )
-from pythainlp.tools import get_full_data_path
-from pythainlp.tools.path import _safe_path_join, get_pythainlp_data_path
+from pythainlp.tools import get_full_data_path, safe_path_join
+from pythainlp.tools.path import get_pythainlp_data_path
 
 
 class SecurityTestCase(unittest.TestCase):
@@ -208,27 +208,27 @@ def test_get_full_data_path_rejects_multiple_traversal(self):
         self.assertIn("path traversal", str(ctx.exception).lower())
 
     def test_safe_path_join_bundled_corpus_safe(self):
-        """Test _safe_path_join with corpus_path() base accepts safe filenames."""
-        result = _safe_path_join(corpus_path(), "negations_th.txt")
+        """Test safe_path_join with corpus_path() base accepts safe filenames."""
+        result = safe_path_join(corpus_path(), "negations_th.txt")
         self.assertTrue(_is_within_directory(corpus_path(), result))
 
     def test_safe_path_join_bundled_corpus_rejects_traversal(self):
-        """Test _safe_path_join with corpus_path() base rejects traversal."""
+        """Test safe_path_join with corpus_path() base rejects traversal."""
         with self.assertRaises(ValueError) as ctx:
-            _safe_path_join(corpus_path(), "../../etc/passwd")
+            safe_path_join(corpus_path(), "../../etc/passwd")
         self.assertIn("path traversal", str(ctx.exception).lower())
 
     def test_safe_path_join_with_tmpdir_safe(self):
-        """Test _safe_path_join accepts safe sub-paths within a temp directory."""
+        """Test safe_path_join accepts safe sub-paths within a temp directory."""
         with tempfile.TemporaryDirectory() as tmpdir:
-            result = _safe_path_join(tmpdir, "model.txt")
+            result = safe_path_join(tmpdir, "model.txt")
             self.assertTrue(_is_within_directory(tmpdir, result))
 
     def test_safe_path_join_with_tmpdir_rejects_traversal(self):
-        """Test _safe_path_join rejects traversal escape from a temp directory."""
+        """Test safe_path_join rejects traversal escape from a temp directory."""
         with tempfile.TemporaryDirectory() as tmpdir:
             with self.assertRaises(ValueError) as ctx:
-                _safe_path_join(tmpdir, "../../etc/passwd")
+                safe_path_join(tmpdir, "../../etc/passwd")
             self.assertIn("path traversal", str(ctx.exception).lower())
 
 

From 10e981c0f9fa3660d9487c0b2de00600ba2d31f4 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 19 Mar 2026 08:43:37 +0000
Subject: [PATCH 06/15] fix: remove duplicate FileNotFoundError guard; update
 CHANGELOG entry

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 CHANGELOG.md                        | 3 +--
 pythainlp/tag/wangchanberta_onnx.py | 9 +--------
 2 files changed, 2 insertions(+), 10 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 5c88643fd..282e47a54 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -28,8 +28,7 @@ This release focuses on security issues related to corpus file loading.
 - w2p: Use npz model instead of pickle (#1328)
 - Fix path traversal vulnerabilities: validate that paths constructed from
   external input stay within their expected base directory;
-  `get_full_data_path()`, `path_pythainlp_corpus()`, and
-  `get_path_folder_corpus()` now raise `ValueError` on traversal attempts
+  `get_full_data_path()` now raises `ValueError` on traversal attempts
 
 ## [5.3.0] - 2026-03-10
 
diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index e4fa34a27..f1c9f7fe1 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -50,14 +50,7 @@ def __init__(
         )
         _corpus_base = get_corpus_path(self.model_name, self.model_version)
         if not _corpus_base:
-            raise FileNotFoundError(
-                f"corpus-not-found name={self.model_name!r} "
-                f"version={self.model_version!r}\n"
-                f"  Corpus '{self.model_name}' "
-                f"(version {self.model_version}) not found.\n"
-                f"    Python: pythainlp.corpus.download('{self.model_name}')\n"
-                f"    CLI:    thainlp data get {self.model_name}"
-            )
+            raise FileNotFoundError(self.model_name)
         self.session = InferenceSession(
             safe_path_join(_corpus_base, file_onnx),
             sess_options=self.options,

From cad6cdd34c60b11963d074e3169375fc1a58327e Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Thu, 19 Mar 2026 09:01:42 +0000
Subject: [PATCH 07/15] style: apply black formatting; fix flake8 E303 in
 test_security.py

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/tag/wangchanberta_onnx.py | 30 +++++++++--------------------
 tests/core/test_security.py         |  1 -
 2 files changed, 9 insertions(+), 22 deletions(-)

diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index f1c9f7fe1..d7a6aeed0 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -45,9 +45,7 @@ def __init__(
         self.model_name = model_name
         self.model_version = model_version
         self.options = SessionOptions()
-        self.options.graph_optimization_level = (
-            GraphOptimizationLevel.ORT_ENABLE_ALL
-        )
+        self.options.graph_optimization_level = GraphOptimizationLevel.ORT_ENABLE_ALL
         _corpus_base = get_corpus_path(self.model_name, self.model_version)
         if not _corpus_base:
             raise FileNotFoundError(self.model_name)
@@ -59,9 +57,7 @@ def __init__(
         self.session.disable_fallback()
         self.outputs_name = self.session.get_outputs()[0].name
         self.sp = spm.SentencePieceProcessor(
-            model_file=safe_path_join(
-                _corpus_base, "sentencepiece.bpe.model"
-            )
+            model_file=safe_path_join(_corpus_base, "sentencepiece.bpe.model")
         )
         with open(
             safe_path_join(_corpus_base, "config.json"),
@@ -76,9 +72,7 @@ def build_tokenizer(self, sent: str) -> dict[str, "np.ndarray"]:
         _t = [5] + [i + 4 for i in self.sp.encode(sent)] + [6]
         model_inputs = {}
         model_inputs["input_ids"] = np.array([_t], dtype=np.int64)
-        model_inputs["attention_mask"] = np.array(
-            [[1] * len(_t)], dtype=np.int64
-        )
+        model_inputs["attention_mask"] = np.array([[1] * len(_t)], dtype=np.int64)
         return model_inputs
 
     def postprocess(self, logits_data: "np.ndarray") -> "np.ndarray":
@@ -90,9 +84,7 @@ def postprocess(self, logits_data: "np.ndarray") -> "np.ndarray":
         scores = shifted_exp / shifted_exp.sum(axis=-1, keepdims=True)
         return scores
 
-    def clean_output(
-        self, list_text: list[tuple[str, str]]
-    ) -> list[tuple[str, str]]:
+    def clean_output(self, list_text: list[tuple[str, str]]) -> list[tuple[str, str]]:
         return list_text
 
     def totag(self, post: np.ndarray, sent: str) -> list[tuple[str, str]]:
@@ -102,25 +94,21 @@ def totag(self, post: np.ndarray, sent: str) -> list[tuple[str, str]]:
             tag.append(
                 (
                     _s[i],
-                    self.id2tag[
-                        str(list(post[i + 1]).index(max(list(post[i + 1]))))
-                    ],
+                    self.id2tag[str(list(post[i + 1]).index(max(list(post[i + 1]))))],
                 )
             )
         return tag
 
-    def _config(
-        self, list_ner: list[tuple[str, str]]
-    ) -> list[tuple[str, str]]:
+    def _config(self, list_ner: list[tuple[str, str]]) -> list[tuple[str, str]]:
         return list_ner
 
     def get_ner(
         self, text: str, tag: bool = False
     ) -> Union[str, list[tuple[str, str]]]:
         self._s: dict[str, "np.ndarray"] = self.build_tokenizer(text)
-        logits = self.session.run(
-            output_names=[self.outputs_name], input_feed=self._s
-        )[0]
+        logits = self.session.run(output_names=[self.outputs_name], input_feed=self._s)[
+            0
+        ]
         _tag = self.clean_output(self.totag(self.postprocess(logits), text))
         if tag:
             _tag = self._config(_tag)
diff --git a/tests/core/test_security.py b/tests/core/test_security.py
index 8f9b48d0c..07095b942 100644
--- a/tests/core/test_security.py
+++ b/tests/core/test_security.py
@@ -189,7 +189,6 @@ def test_is_within_directory_with_symlinks(self):
             # in the _safe_extract_tar and _safe_extract_zip functions,
             # which check where symlinks actually point to.
 
-
     def test_get_full_data_path_safe(self):
         """Test that get_full_data_path returns a path within the data directory."""
         result = get_full_data_path("ttc_freq.txt")

From 57d2839e2d55efcbb3124755e48d53a1e00b3784 Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:10:16 +0000
Subject: [PATCH 08/15] Update pythainlp/tag/wangchanberta_onnx.py

---
 pythainlp/tag/wangchanberta_onnx.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index d7a6aeed0..ec56fabbe 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -45,7 +45,9 @@ def __init__(
         self.model_name = model_name
         self.model_version = model_version
         self.options = SessionOptions()
-        self.options.graph_optimization_level = GraphOptimizationLevel.ORT_ENABLE_ALL
+        self.options.graph_optimization_level = (
+            GraphOptimizationLevel.ORT_ENABLE_ALL
+        )
         _corpus_base = get_corpus_path(self.model_name, self.model_version)
         if not _corpus_base:
             raise FileNotFoundError(self.model_name)

From cde810ffff127ce904c5523f662ec5f2d52f3c36 Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:10:23 +0000
Subject: [PATCH 09/15] Update pythainlp/tag/wangchanberta_onnx.py

---
 pythainlp/tag/wangchanberta_onnx.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index ec56fabbe..c4c561be2 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -74,7 +74,9 @@ def build_tokenizer(self, sent: str) -> dict[str, "np.ndarray"]:
         _t = [5] + [i + 4 for i in self.sp.encode(sent)] + [6]
         model_inputs = {}
         model_inputs["input_ids"] = np.array([_t], dtype=np.int64)
-        model_inputs["attention_mask"] = np.array([[1] * len(_t)], dtype=np.int64)
+        model_inputs["attention_mask"] = np.array(
+            [[1] * len(_t)], dtype=np.int64
+        )
         return model_inputs
 
     def postprocess(self, logits_data: "np.ndarray") -> "np.ndarray":

From a61e4e6091fb3f76292b8d9e60b723c07eca97b5 Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:10:32 +0000
Subject: [PATCH 10/15] Update pythainlp/tag/wangchanberta_onnx.py

---
 pythainlp/tag/wangchanberta_onnx.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index c4c561be2..958174024 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -88,7 +88,9 @@ def postprocess(self, logits_data: "np.ndarray") -> "np.ndarray":
         scores = shifted_exp / shifted_exp.sum(axis=-1, keepdims=True)
         return scores
 
-    def clean_output(self, list_text: list[tuple[str, str]]) -> list[tuple[str, str]]:
+    def clean_output(
+        self, list_text: list[tuple[str, str]]
+    ) -> list[tuple[str, str]]:
         return list_text
 
     def totag(self, post: np.ndarray, sent: str) -> list[tuple[str, str]]:

From 504216d5aafb5551646a8f92b2c379c6ea2bc9ad Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:10:38 +0000
Subject: [PATCH 11/15] Update pythainlp/tag/wangchanberta_onnx.py

---
 pythainlp/tag/wangchanberta_onnx.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index 958174024..d59c2e20a 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -100,7 +100,9 @@ def totag(self, post: np.ndarray, sent: str) -> list[tuple[str, str]]:
             tag.append(
                 (
                     _s[i],
-                    self.id2tag[str(list(post[i + 1]).index(max(list(post[i + 1]))))],
+                    self.id2tag[
+                        str(list(post[i + 1]).index(max(list(post[i + 1]))))
+                    ],
                 )
             )
         return tag

From d0ccffbafb1a73f449c85e1c889602defdcd8d6a Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:10:46 +0000
Subject: [PATCH 12/15] Update pythainlp/tag/wangchanberta_onnx.py

---
 pythainlp/tag/wangchanberta_onnx.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index d59c2e20a..e7d5ed0f3 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -107,7 +107,9 @@ def totag(self, post: np.ndarray, sent: str) -> list[tuple[str, str]]:
             )
         return tag
 
-    def _config(self, list_ner: list[tuple[str, str]]) -> list[tuple[str, str]]:
+    def _config(
+        self, list_ner: list[tuple[str, str]]
+    ) -> list[tuple[str, str]]:
         return list_ner
 
     def get_ner(

From c2881ae20b81ed7fa0cc944ab332b008dacfaddb Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:10:53 +0000
Subject: [PATCH 13/15] Update pythainlp/tag/wangchanberta_onnx.py

---
 pythainlp/tag/wangchanberta_onnx.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/pythainlp/tag/wangchanberta_onnx.py b/pythainlp/tag/wangchanberta_onnx.py
index e7d5ed0f3..cf62104c0 100644
--- a/pythainlp/tag/wangchanberta_onnx.py
+++ b/pythainlp/tag/wangchanberta_onnx.py
@@ -116,9 +116,9 @@ def get_ner(
         self, text: str, tag: bool = False
     ) -> Union[str, list[tuple[str, str]]]:
         self._s: dict[str, "np.ndarray"] = self.build_tokenizer(text)
-        logits = self.session.run(output_names=[self.outputs_name], input_feed=self._s)[
-            0
-        ]
+        logits = self.session.run(
+            output_names=[self.outputs_name], input_feed=self._s
+        )[0]
         _tag = self.clean_output(self.totag(self.postprocess(logits), text))
         if tag:
             _tag = self._config(_tag)

From c01ad3300ff7d5f7096785810afbb35d403540a1 Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:13:12 +0000
Subject: [PATCH 14/15] Update CHANGELOG for release 5.3.2

This release addresses security issues related to path traversal by validating that paths stay within their expected base directory.
---
 CHANGELOG.md | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 282e47a54..bb9b8e119 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -15,7 +15,16 @@ and this project adheres to
 [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 - Full release notes: <https://github.com/PyThaiNLP/pythainlp/releases>
-- Commit history: <https://github.com/PyThaiNLP/pythainlp/compare/v5.3.0...v5.3.1>
+- Commit history: <https://github.com/PyThaiNLP/pythainlp/compare/v5.3.1...v5.3.2>
+
+## [5.3.2] - 2026-03-20
+
+This release focuses on security issues related to path traversal.
+
+### Security
+
+- Prevent path traversal: validate that paths stay within
+  their expected base directory (#1342)
 
 ## [5.3.1] - 2026-03-14
 
@@ -26,9 +35,6 @@ This release focuses on security issues related to corpus file loading.
 - thai2fit: Use JSON model instead of pickle (#1325)
 - Defensive corpus loading: validate fields before processing (#1327)
 - w2p: Use npz model instead of pickle (#1328)
-- Fix path traversal vulnerabilities: validate that paths constructed from
-  external input stay within their expected base directory;
-  `get_full_data_path()` now raises `ValueError` on traversal attempts
 
 ## [5.3.0] - 2026-03-10
 
@@ -215,6 +221,7 @@ The minimum requirement is now Python 3.9.
 
 - See <https://github.com/PyThaiNLP/pythainlp/releases/tag/v5.0.0>
 
+[5.3.2]: https://github.com/PyThaiNLP/pythainlp/compare/v5.3.1...v5.3.2
 [5.3.1]: https://github.com/PyThaiNLP/pythainlp/compare/v5.3.0...v5.3.1
 [5.3.0]: https://github.com/PyThaiNLP/pythainlp/compare/v5.2.0...v5.3.0
 [5.2.0]: https://github.com/PyThaiNLP/pythainlp/compare/v5.1.2...v5.2.0

From 54d11e606cb54babe3214dd640793615a77ffb82 Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Thu, 19 Mar 2026 09:16:23 +0000
Subject: [PATCH 15/15] Update AGENTS.md about path join

---
 AGENTS.md | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/AGENTS.md b/AGENTS.md
index ab126798e..6d225a119 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -6,7 +6,7 @@
       <https://github.com/PyThaiNLP/pythainlp/blob/dev/tests/README.md>.
       The document list test categories, their dependency sets,
       and test naming conventions.
-- [ ] Use reStructuredText for docstring (PEP 287), targetting Sphinx.
+- [ ] Use reStructuredText for docstring (PEP 287), targeting Sphinx.
 - [ ] When possible, follow NLTK established convention of submodule
       name (tend to be a verb or a generic noun), function name, and
       configuration. Communicate this to the users during code review.
@@ -20,7 +20,7 @@
       the repo. Read its usage and information it generates at
       <https://github.com/PyThaiNLP/pythainlp/blob/dev/build_tools/analysis/README.md>.
       Mind that the analyzer can create false positives,
-      please refer to Python tyep specification when in doubt.
+      please refer to Python type specification when in doubt.
 - [ ] Complete type annotations for function, method, class, variable, etc.
       Maintain near-100% type annotation coverage.
 - [ ] Add tests for new functionality or behavior.
@@ -31,6 +31,9 @@
 - [ ] Major changes should be logged in the change log at
       <https://github.com/PyThaiNLP/pythainlp/blob/dev/CHANGELOG.md>.
       Provide issue number or PR number if available.
+- [ ] Do not use os.path.join();
+      always use pythainlp.tools.safe_path_join() instead,
+      to prevent path traversal vulnerabilities (CWE-22).
 
 ## Project contribution guidelines