git-vuln-finder

Finding potential software vulnerabilities from git commit messages. The output format is a JSON with the associated commit which could contain a fix regarding a software vulnerability. The search is based on a set of regular expressions against the commit messages only. If CVE IDs are present, those are added automatically in the output. The input can be any git repositories or a GH archive source.

Requirements

• jq (sudo apt install jq)
• Python poetry

Installation

Use it as a library

git-vuln-finder can be install with poetry. If you don't have poetry installed, you can do the following curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python.

$ poetry install
$ poetry shell
$ git-vuln-finder -h

You can also use pip. Then just import it:

Python 3.8.0 (default, Dec 11 2019, 21:43:13)
[GCC 9.2.1 20191008] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from git_vuln_finder import find
>>> all_potential_vulnerabilities, all_cve_found, found = find("~/git/curl")

>>> [commit for commit, summary in all_potential_vulnerabilities.items() if summary['state'] == 'cve-assigned']
['9069838b30fb3b48af0123e39f664cea683254a5', 'facb0e4662415b5f28163e853dc6742ac5fafb3d',
... snap ...
 '8a75dbeb2305297640453029b7905ef51b87e8dd', '1dc43de0dccc2ea7da6dddb7b98f8d7dcf323914', '192c4f788d48f82c03e9cef40013f34370e90737', '2eb8dcf26cb37f09cffe26909a646e702dbcab66', 'fa1ae0abcde5df8d0b3283299e3f246bedf7692c', 'c11c30a8c8d727dcf5634fa0cc6ee0b4b77ddc3d', '75ca568fa1c19de4c5358fed246686de8467c238', 'a20daf90e358c1476a325ea665d533f7a27e3364', '042cc1f69ec0878f542667cb684378869f859911']

 >>> print(json.dumps(all_potential_vulnerabilities['9069838b30fb3b48af0123e39f664cea683254a5'], sort_keys=True, indent=4, separators=(",", ": ")))
 {
     "author": "Daniel Stenberg",
     "author-email": "[email protected]",
     "authored_date": 1567544372,
     "branches": [
         "master"
     ],
     "commit-id": "9069838b30fb3b48af0123e39f664cea683254a5",
     "committed_date": 1568009674,
     "cve": [
         "CVE-2019-5481",
         "CVE-2019-5481"
     ],
     "language": "en",
     "message": "security:read_data fix bad realloc()\n\n... that could end up a double-free\n\nCVE-2019-5481\nBug: https://curl.haxx.se/docs/CVE-2019-5481.html\n",
     "origin": "https://github.com/curl/curl.git",
     "origin-github-api": "https://api.github.com/repos///github.com/curl/curl/commits/9069838b30fb3b48af0123e39f664cea683254a5",
     "pattern-matches": [
         "double-free"
     ],
     "pattern-selected": "(?i)(double[-| ]free|buffer overflow|double free|race[-| ]condition)",
     "state": "cve-assigned",
     "stats": {
         "deletions": 4,
         "files": 1,
         "insertions": 2,
         "lines": 6
     },
     "summary": "security:read_data fix bad realloc()",
     "tags": []
 }

Use it as a command line tool

$ git clone https://github.com/cve-search/git-vuln-finder.git
$ cd https://github.com/cve-search/git-vuln-finder.git
$ pip install .
$ git-vuln-finder --help

You can also use pip. pipx installs scripts (system wide available) provided by Python packages into separate virtualenvs to shield them from your system and each other.

Usage

usage: git-vuln-finder [-h] [-v] [-r R] [-o O] [-s S] [-p P] [-c] [-t] [-gh GH]

Finding potential software vulnerabilities from git commit messages.

optional arguments:
  -h, --help  show this help message and exit
  -v          increase output verbosity
  -r R        git repository to analyse
  -o O        Output format: [json]
  -s S        State of the commit found
  -p P        Matching pattern to use: [vulnpatterns, cryptopatterns, cpatterns] - the pattern 'all' is used to match all the patterns at once.
  -c          output only a list of the CVE pattern found in commit messages (disable by default)
  -t          Include tags matching a specific commit
  -gh GH      special option for gharchive, pass a file containing a PushEvent in JSON format

More info: https://github.com/cve-search/git-vuln-finder

Patterns

git-vuln-finder comes with 3 default patterns which can be selected to find the potential vulnerabilities described in the commit messages such as:

• vulnpatterns is a generic vulnerability pattern especially targeting web application and generic security commit message. Based on an academic paper.
• cryptopatterns is a vulnerability pattern for cryptographic errors mentioned in commit messages.
• cpatterns is a set of standard vulnerability patterns see for C/C++-like languages.

A sample partial output from Curl git repository