fpvs 0.3.0

Fast Python Vulnerability Scanner

Fast Python Vulnerability Scanner

Use the GitLab Advisory Database to do Python Vulnerability Scanning.

This looks quite similar to what Gitlab's Gemnasium Dependency Scanning Analyzer does, but comes with some differences:

• No Docker container is included in this project. Bring your own or work locally.
• Results are printed on stdout, and are directly readable.
• Works on a directory of already vendored wheels rather than doing pip install -r requirements.txt in some half-baked attempt to determine the results of such a call.

Why a single command, rather than a full blown Docker container?

1. You should generally ask the opposite question. Now get off my lawn.
2. Ability to trivially run from your local development environment.
3. No working around the limitations of missing header files when trying to get pip install some-package-without-wheel-but-with-c-bindings to work.
4. Fast: just reuse already-vendored packages (from some cache or artifact) rather than doing yet another pip install -r requirements.txt in each job of your pipeline.

usage: fpvs-scan [-h] [--wheels-path WHEELS_PATH] [--gemnasium-db-path GEMNASIUM_DB_PATH] [--version] [--verbose] [--no-invalid-specifiers]

Example usage:

# 1. install the wheels to be scanned
pip install wheel
pip wheel requests==2.0.1 --wheel-dir=vendor  # this is a bad package on purpose, to show off what FPVS does

# 2. get fpvs, the vulnerability database and do the scanning:
pip install fpvs
git clone [e-mail]:gitlab-org/security-products/gemnasium-db.git
fpvs-scan --verbose

In typical real-world usage, part 1 of the example above would already be executed in some other part of your pipeline or development flow, and would have become an artifact / just live on your machine. This is what makes FPVS fast: it doesn't do slow stuff that you did already.

Example output:

fpvs-scan  --verbose
Checking wheels in vendor against gemnasium-db
SCANNING requests-2.0.1-py2.py3-none-any.whl
ADVISORY pypi/requests/GMS-2012-3.yml: 2.0.1 against <0.12.0 OK
ADVISORY pypi/requests/CVE-2014-1830.yml: 2.0.1 against <=2.2.1 FAIL
ADVISORY pypi/requests/CVE-2013-2099.yml: 2.0.1 against <=1.0.3 OK
ADVISORY pypi/requests/CVE-2015-2296.yml: 2.0.1 against <2.6.0 FAIL
ADVISORY pypi/requests/CVE-2014-1829.yml: 2.0.1 against <=2.2.1 FAIL
ADVISORY pypi/requests/CVE-2018-18074.yml: 2.0.1 against <2.20.0 FAIL

requests==2.0.1: 4 advisories

1. Information Exposure
Requests (aka python-requests) allows remote servers to obtain sensitive information by reading the Proxy-Authorization header in a redirected request.
CVE-2014-1830

2. Session fixation in resolve_redirects()
The `resolve_redirects()` function in `sessions.py` allows a remote, user-assisted attacker to conduct a session fixation attack. This flaw exists because the application, when establishing a new session, does not invalidate an existing session identifier and assign a new one. With a specially crafted request fixating the session identifier, a context-dependent attacker can ensure a user authenticates with the known session identifier, allowing the session to be subsequently hijacked.
CVE-2015-2296

3. Information Exposure
Requests (aka python-requests) allows remote servers to obtain a netrc password by reading the Authorization header in a redirected request.
CVE-2014-1829

4. Information exposure in HTTP headers
The Requests package for Python sends an HTTP Authorization header to an HTTP URI upon receiving a same-hostname https-to-http redirect, which makes it easier for remote attackers to discover credentials by sniffing the network.
CVE-2018-18074

Solutions (for requests==2.0.1)
Upgrade to version 2.20.0 or above.
Upgrade to version 2.3.0 or above.
Upgrade to version 2.6.0 or above.

FAILURE: Found 1 unsafe packages

Or if you don't like reading:

$ fpvs-scan
requests==2.0.1: 4 advisories
Upgrade to version 2.20.0 or above.
Upgrade to version 2.3.0 or above.
Upgrade to version 2.6.0 or above.

FAILURE: Found 1 unsafe packages

The GitLab Advisory Database may occasionally contain invalid version specifiers. If you want fpvs to fail in that scenario, run it with --no-invalid-specifiers.

Note that the GitLab Advisory Database has a licence that is separate from the FPVS.