This pull request introduces a new CodeQL query for detecting prompt injection vulnerabilities in Python code targeting AI prompting APIs such as agents and openai. The changes includes a new experimental query, new taint flow and type models, a customizable dataflow configuration, documentation, and comprehensive test coverage.
See https://docs.python.org/3/library/compression.zstd.html for
information about this library.
As far as I can tell, the `zstd` library is not vulnerable to things
like ZipSlip, but it _could_ be vulnerable to a decompression bomb
attack, so I extended those models accordingly.
Fixed 73 .ql query files where the @name metadata contained an ending period.
This ensures consistency with the CodeQL query metadata style guidelines.
This PR adds a query to detect a Cross Origin Resource Sharing(CORS) policy bypass due to an incorrect check.
This PR attempts to detect the vulnerability pattern found in CVE-2022-3457
```python
if request.method in ['POST', 'PUT', 'PATCH', 'DELETE']:
origin = request.headers.get('Origin', None)
if origin and not origin.startswith(request.base):
raise cherrypy.HTTPError(403, 'Unexpected Origin header')
```
In this case, a value obtained from a header is compared using `startswith` call. This comparision is easily bypassed resulting in a CORS bypass. Given that similar bugs have been found in other languages as well, I think this PR would be a great addition to the exisitng python query pack.
The databases for CVE-2022-3457 can be downloaded from
```
https://filetransfer.io/data-package/i4Mfepls#linkhttps://file.io/V67T4SSgmExF
```
Js2Py is a Javascript to Python translation library written in Python. It allows users to invoke JavaScript code directly from Python.
The Js2Py interpreter by default exposes the entire standard library to it's users. This can lead to security issues if a malicious input were directly.
This PR includes a CodeQL query along with a qhelp and testcases to detect cases where an untrusted input flows to an Js2Py eval call.
This query successfully detects CVE-2023-0297 in `pyload/pyload`along with it's fix.
The databases can be downloaded from the links bellow.
```
https://file.io/qrMEjSJJoTq1https://filetransfer.io/data-package/a02eab7V#link
```
