Vulnerabilities

Vulnerabilities refer to potential security flaws or weaknesses in PyAnsys packages. Here are some examples of common vulnerabilities:

• Insufficient input validation or insecure handling of file paths leading to unauthorized access

• Insecure code and command injection into the app

• Remote code execution due to deserialization of untrusted code

• Disclosure of sensitive information due to weak exception handling

Vulnerability sources

• Vulnerabilities from PyAnsys library code: Maintainers are responsible for deciding whether to address vulnerabilities. The priority of vulnerabilities can be escalated internally if they represent a roadblock for usage. The Ansys business unit in charge of the project should handle vulnerabilities on a case-by-case basis.

• Vulnerabilities from external package dependencies: When vulnerabilities exist in external packages used by PyAnsys libraries, such as NumPy or Matplotlib, Ansys should not address these vulnerabilities directly. Instead, raise an issue on the open source repository, pointing out the vulnerability and linking the applicable CVE. At most, be mindful of deprecated packages and functions because they may not receive security patches and might introduce vulnerabilities into your codebase.

Vulnerability discovery and tracking

Leverage available security tools like dependency scanners or static analyzers (such as PyUp, Safety, or Bandit) to automatically detect and remediate security vulnerabilities in Python packages and dependencies.

The PyAnsys ecosystem has implemented automated mechanisms to track and report vulnerabilities in the codebase. These tools are intended to be integrated into the CI/CD workflow of the repositories.

The PyAnsys Core team has come up with a GitHub Action that can be used to automatically check for vulnerabilities in the codebase. This action is based on the following tools:

• Safety : Checks installed dependencies for known security vulnerabilities.

• Bandit: Attempts to find common security issues in Python code.

Safety addresses external dependencies, while Bandit focuses on the codebase itself. Both tools have been integrated into the ansys/actions/check-vulnerabilities action.

Note

The tools selected for the action are not definitive. The action can be modified to include other tools or to use different versions of the tools in the future.

For third-party packages, the PyAnsys Core team has listed a set of excluded advisories so that the action does not fail. This is done to avoid false positives and to ensure that the action does not block the CI/CD pipeline unnecessarily. You can find the list of excluded advisories in the check-vulnerabilities action documentation.

For potential vulnerabilities in the codebase, repositories can configure Bandit to ignore specific advisories. This can be due to the code not being ready yet to be fixed or that the advisory is not relevant to the codebase. However, it is important to note that ignoring advisories should be done with caution, and developers should be aware of the potential risks involved. Furthermore, repository maintainers should document the reasons for ignoring advisories and ensure that they are regularly reviewed to determine if they can be addressed.

Note

An example on how to document the ignored advisories can be found in the PyACP security considerations documentation page. This should be taken as a reference for documenting ignored advisories in other repositories.

Warning

Testing the action locally before enabling it in the CI/CD workflow is recommended. Information on how to do this can be found in the check-vulnerabilities action documentation.

Vulnerability remediation and reporting

When a vulnerability is detected, the action fails and reports the vulnerabilities found in the codebase. Following that, these vulnerabilities are reported as draft security advisories in the repository’s Security tab. Maintainers are then responsible for reviewing the advisories and deciding whether to address them or not. These advisories are monitored by the PyAnsys Core team and are escalated internally if they represent a roadblock for usage.

Repositories should also have a process in place to handle vulnerabilities that are reported by users or other developers. For that purpose, a SECURITY.md file should be created in the root of the repository. This file should contain information on how to report vulnerabilities and the process for handling them.

Here is an example of a SECURITY.md file:

## Reporting a vulnerability

> [!CAUTION]
> Do not use GitHub issues to report any security vulnerabilities.

If you detect a vulnerability, contact the [PyAnsys Core team](mailto:pyansys.core@ansys.com),
mentioning the repository and the details of your finding. The team will address it as soon as possible.

Provide the PyAnsys Core team with this information:

- Any specific configuration settings needed to reproduce the problem
- Step-by-step guidance to reproduce the problem
- The exact location of the problematic source code, including tag, branch, commit, or a direct URL
- The potential consequences of the vulnerability, along with a description of how an attacker could take advantage of the issue

Vulnerability disclosure

When a vulnerability is detected and a decision is made to address it, the repository maintainers should create a private fork of the repository and create a pull request with the fix. Information on how to create such a temporary fork to resolve a vulnerability can be found in Github’s documentation. When opened, the pull request should be reviewed in depth and include tests to ensure that the vulnerability is fixed. Once the pull request is merged, the repository maintainers should create a new release with the fix and update the changelog accordingly.

The release should be tagged with a new version number where the patch value has been incremented, and the changelog should include a note about the vulnerability and the fix. The note should include this information:

• The CVE number of the vulnerability (if applicable)

• A description of the vulnerability and its potential consequences

• A description of the fix and how it addresses the vulnerability

• A link to the pull request that fixed the vulnerability

Additionally, the security advisory should be published on the repository’s Security tab. This advisory should include the same information as the changelog note, in other words the CVE number, the date of the advisory, and the status of the advisory (such as published or withdrawn).

A reference of a published security advisory can be found here: PyAnsys Geometry subprocess advisory. This advisory was published in the PyAnsys Geometry repository and includes information about a vulnerability in which users could execute arbitrary code on the system by using one of this library’s functions.

Ensuring compliance across the PyAnsys ecosystem

The PyAnsys Core team is responsible for ensuring that the ansys/actions/check-vulnerabilities action is up to date and that it is being used in all PyAnsys repositories considered as libraries (that is, Python packages shipped to PyPI). Repository maintainers are responsible for ensuring that the action is implemented correctly and that the results are reviewed regularly.

Addressing common vulnerabilities in python libraries and applications

When developing Python applications, it is essential to be aware of common vulnerabilities that can occur in the codebase. These vulnerabilities can lead to security risks, data breaches, and other serious issues.

The Bandit tool provides a blacklist of known vulnerable functions and methods that should not be used in Python code. Using these functions can lead to security vulnerabilities and should be avoided. Refer to the blacklists Bandit documentation for detailed information on Bandit tool outputs.

Bandit blacklist

The Bandit tool provides a blacklist of known vulnerable functions and methods that should not be used in Python code. Using these functions can lead to security vulnerabilities and should be avoided.

Address each requested changes proposed by Bandit to ensure that your code is secure. You can find information on how to improve your code in the blacklists Bandit documentation.

subprocess command injection

The subprocess module can be vulnerable to command injection if user input is not properly sanitized. This can lead to arbitrary command execution, which is a significant security risk.

To mitigate this risk, you should:

• avoid using the subprocess module to execute shell commands with user input, as it can lead to command injection vulnerabilities.

• if the previous point is not possible, you need to disable the shell=True argument in subprocess.run() or similar functions, as it allows for shell injection attacks.

By removing the shell=True argument, a list is needed to pass the command and its arguments directly, which is safer. This way, user input is not executed as a shell command, and the risk of command injection is significantly reduced.

Risk of subprocess command injection

import subprocess

user_input = "malicious_command; rm -rf /"  # User input that could be malicious
subprocess.run(f"echo {user_input}", shell=True)  # Vulnerable to command injection

Reduced risk of subprocess command injection

import subprocess

user_input = "malicious_command; rm -rf /"  # User input that could be malicious
# Removing shell=True and using a list
subprocess.run(["echo", user_input])  # User input is not executed as a shell command

Note

Bandit warning remains even after deactivating the shell=True argument. If you are sure that the command is safe, you can ignore the Bandit warning. Please check the Ignore Bandit warnings section for more information on how to do so.

try except continue statements

Using try except continue statements can lead to silent failures, making it difficult to debug issues and potentially allowing vulnerabilities to go unnoticed. Instead, you should handle exceptions explicitly and log or raise them as needed.

try except continue without handling exceptions

try:
    risky_operation()  # Some code that might raise an exception
except:
    continue  # This will silently ignore all the exceptions and continue execution

try except continue with explicit exception handling

try:
    risky_operation()
except SpecificException as e:
    continue  # Handle specific exceptions and continue
except AnotherSpecificException as e:
    log_error(e)  # Log the error for debugging
    raise  # Raise the exception to notify the caller

requests.get() without timeout

Using requests.get() without a timeout can lead to hanging requests, which can be exploited by attackers to cause denial of service (DoS) conditions. Always specify a timeout value to prevent this issue.

requests.get() without timeout

import requests

response = requests.get("https://example.com")  # No timeout specified

requests.get() with timeout

import requests

response = requests.get("https://example.com", timeout=5)  # Timeout set to 5 seconds

random insecure functions

Using insecure functions from the random module can lead to predictable random number generation, which can be exploited by attackers. Instead, use the secrets module, which provides a secure way to generate random numbers.

Insecure random functions

import random

random_number = random.randint(1, 100)  # Predictable random number generation
random_letter = random.choice(["a", "b", "c"])  # Predictable choice from a list

Secure random functions

import secrets

secure_random_number = secrets.randbelow(100)  # Secure random number generation
secure_random_letter = secrets.choice(["a", "b", "c"])  # Secure choice from a list

Ignore Bandit warnings

In-line comment

When using Bandit, you may encounter warnings that you believe are not relevant to your codebase or that you have already addressed. In such cases, you can ignore specific Bandit warnings by adding a comment to the end of the line that triggers the warning. The comment should be in the format # nosec <warning_id>, where <warning_id> is the ID of the warning you want to ignore.

When you ignore a Bandit warning, it is essential to provide a clear comment explaining why the warning is being ignored. This helps maintainers and other developers understand the context and rationale behind the decision.

For example, to ignore the B404 warning, you would add # nosec B404 to the end of the line:

# Subprocess is needed to start the backend. But
# the input is controlled by the library. Excluding bandit check.
import subprocess  # nosec B404

Warning

Please note that ignoring Bandit warnings should be done with caution, and you should ensure that the code is safe and does not introduce any security risks. It is recommended to review the bandit documentation for more information on each warning and the potential risks involved.

Security considerations file

In addition to ignoring specific Bandit warnings, it is a good practice to document the ignored advisories in a dedicated file. You can find an example of such a file in the PyACP security considerations documentation page. This way, you can provide to the users a clear overview of the vulnerabilities that need to be taken into account when using the library.

Addressing common vulnerabilities in Github Actions

Vulnerabilities can exist in continuous integration (CI) pipelines just as they can in a codebase. To reduce the risk of security breaches and supply chain attacks, it is important to secure your GitHub Actions workflows against known vulnerabilities.

zizmor is a static analysis tool that audits GitHub Actions CI/CD setups. It detects common vulnerabilities and, in some cases, can automatically fix them. For detailed information about the rules that zizmor applies when auditing workflows, see zizmor audit rules.

Auditing CI/CD setups in the PyAnsys ecosystem

For PyAnsys ecosystem projects, the recommended way to audit workflows is to use the ansys/actions/check-actions-security action. The action wraps zizmor and provides additional functionality and configuration tailored to PyAnsys projects. For setup instructions, see the check-actions-security action documentation.

Fixing common issues detected by zizmor

This section describes how to fix common workflow vulnerabilities. For practical examples, see these pull requests with fixes already applied in the PyAnsys ecosystem:

• Ansys actions security fixes 1

• Ansys actions security fixes 2

• Ansys actions security fixes 3

• PyConverter-XML2Py security fixes

For vulnerabilities not listed here, refer to zizmor audit rules for remediation steps. For additional examples of fixes, see the zizmor trophy case.

artipacked

Before

steps:

- name: "Checkout project"
  uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2

After

steps:

- name: "Checkout project"
  uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
  with:
    persist-credentials: false

Note

When you run git commands that require persisted credentials in subsequent steps within the same job, you can ignore this audit finding. For details, see ignoring zizmor results

unpinned-uses

Before

steps:

- name: "Upload distribution artifacts to GitHub artifacts"
  uses: actions/upload-artifact@v4
  with:
    name: ${{ env.LIBRARY_NAME }}-artifacts
    path: ~/${{ env.LIBRARY_NAME }}/dist/

After

steps:

- name: "Upload distribution artifacts to GitHub artifacts"
  uses: actions/upload-artifact@4cec3d8aa04e39d1a68397de0c4cd6fb9dce8ec1 # v4.6.1
  with:
    name: ${{ env.LIBRARY_NAME }}-artifacts
    path: ~/${{ env.LIBRARY_NAME }}/dist/

Note

The ansys/actions/check-actions-security action has a trust-ansys-actions option that allows you to use tags for ansys/actions. When this option is enabled, you only need to pin external actions.

Tip

You can use the pinact tool to automatically pin versions of actions and reusable workflows.

github-env

Before

steps:

- name: "Decompose tag into components"
  shell: bash
  run: |
    if [[ ${{ github.ref_name }} =~ ^v[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
      # Split the tag into its components
      IFS='.' read -ra PARTS <<< "${{ github.ref_name }}"
      echo "V_AND_MAJOR=${PARTS[0]}" >> $GITHUB_ENV
      echo "MINOR=${PARTS[1]}" >> $GITHUB_ENV
      echo "PATCH=${PARTS[2]}" >> $GITHUB_ENV
    else
      echo "Invalid tag format. Expected vX.Y.Z but got ${{ github.ref_name }}"
      exit 1
    fi

- name: "Check tag is valid for current branch"
  shell: bash
  run: |
    # Remove leading "v" from env.X
    V_AND_MAJOR=${{ env.V_AND_MAJOR }}
    MAJOR="${V_AND_MAJOR#v}"
    echo "MAJOR=${MAJOR}" >> $GITHUB_ENV
    if [[ ${{ github.event.base_ref }} != "refs/heads/release/$MAJOR.${{ env.MINOR }}" ]]; then
      echo "::error::Tag ${{ github.ref_name }} does not match branch version. wrong branch."
      exit 1
    fi

- name: "Remove v${{ env.MAJOR }} tag"
  shell: bash
  run: |
    git push --delete origin v${{ env.MAJOR }} && \
      echo "Deleted v${{ env.MAJOR }} tag" || \
      echo "Tag v${{ env.MAJOR }} not found"

- name: "Remove v${{ env.MAJOR }}.${{ env.MINOR }} tag"
  shell: bash
  run: |
    git push --delete origin v${{ env.MAJOR }}.${{ env.MINOR }} && \
      echo "Deleted v${{ env.MAJOR }}.${{ env.MINOR }} tag" || \
      echo "Tag v${{ env.MAJOR }}.${{ env.MINOR }} not found"

- name: "Create new tags"
  shell: bash
  run: |
    git tag v${{ env.MAJOR }}.${{ env.MINOR }}
    git tag v${{ env.MAJOR }}
    git push origin v${{ env.MAJOR }}.${{ env.MINOR }}
    git push origin v${{ env.MAJOR }}

After

steps:

- name: "Decompose tag into components"
  id: tag-components
  shell: bash
  run: |
    if [[ ${{ github.ref_name }} =~ ^v[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
      # Split the tag into its components
      IFS='.' read -ra PARTS <<< "${{ github.ref_name }}"
      echo "V_AND_MAJOR=${PARTS[0]}" >> $GITHUB_OUTPUT
      echo "MINOR=${PARTS[1]}" >> $GITHUB_OUTPUT
      echo "PATCH=${PARTS[2]}" >> $GITHUB_OUTPUT
    else
      echo "Invalid tag format. Expected vX.Y.Z but got ${{ github.ref_name }}"
      exit 1
    fi

- name: "Check tag is valid for current branch"
  id: current-branch-tag-validity
  shell: bash
  env:
    V_AND_MAJOR: ${{ steps.tag-components.outputs.V_AND_MAJOR }}
    MINOR: ${{ steps.tag-components.outputs.MINOR }}
  run: |
    # Remove leading "v" from env.X
    MAJOR="${V_AND_MAJOR#v}"
    echo "MAJOR=${MAJOR}" >> $GITHUB_OUTPUT
    if [[ ${{ github.event.base_ref }} != "refs/heads/release/${MAJOR}.${MINOR}" ]]; then
      echo "::error::Tag ${{ github.ref_name }} does not match branch version. wrong branch."
      exit 1
    fi

- name: "Remove v${{ steps.current-branch-tag-validity.outputs.MAJOR }} tag"
  shell: bash
  env:
    MAJOR: ${{ steps.current-branch-tag-validity.outputs.MAJOR }}
  run: |
    git push --delete origin v${MAJOR} && \
      echo "Deleted v${MAJOR} tag" || \
      echo "Tag v${MAJOR} not found"

- name: "Remove v${{ steps.current-branch-tag-validity.outputs.MAJOR }}.${{ steps.tag-components.outputs.MINOR }} tag"
  shell: bash
  env:
    MAJOR: ${{ steps.current-branch-tag-validity.outputs.MAJOR }}
    MINOR: ${{ steps.tag-components.outputs.MINOR }}
  run: |
    git push --delete origin v${MAJOR}.${MINOR} && \
      echo "Deleted v${MAJOR}.${MINOR} tag" || \
      echo "Tag v${MAJOR}.${MINOR} not found"

- name: "Create new tags"
  shell: bash
  env:
    MAJOR: ${{ steps.current-branch-tag-validity.outputs.MAJOR }}
    MINOR: ${{ steps.tag-components.outputs.MINOR }}
  run: |
    git tag v${MAJOR}.${MINOR}
    git tag v${MAJOR}
    git push origin v${MAJOR}.${MINOR}
    git push origin v${MAJOR}

Note

The trick is to pass state between steps using GITHUB_OUTPUT instead of GITHUB_ENV or GITHUB_PATH. On Windows runners, the same principle applies when running commands in cmd or pwsh; only the syntax differs.

template-injection

Before

name: Example reusable workflow

on:
  workflow_call:
    inputs:
      user-input:
        required: false
        type: string
        default: "user input"

  workflow_dispatch:
    inputs:
      required: false
      type: string
      default: "user input"

jobs:
  example-job:
    name: "Example job"
    runs-on: ubuntu-latest
    steps:

    - name: "Inspect context variables and workflow input"
      run: |
        echo ${{ github.workspace }}
        echo ${{ runner.temp }}
        echo ${{ input.user-input }}

After

name: Example reusable workflow

on:
  workflow_call:
    inputs:
      user-input:
        required: false
        type: string
        default: "user input"

  workflow_dispatch:
    inputs:
      required: false
      type: string
      default: "user input"

jobs:
  example-job:
    name: "Example job"
    runs-on: ubuntu-latest
    steps:

    - name: "Inspect context variables and workflow input"
      env:
        USER_INPUT: ${{ inputs.user-input }}
      run: |
        echo ${USER_INPUT}
        echo ${RUNNER_TEMP}
        echo ${GITHUB_WORKSPACE}

Note

Notice that RUNNER_TEMP and GITHUB_WORKSPACE were not explicitly set in the env block. Some GitHub context variables automatically map to environment variables, such as runner.temp to RUNNER_TEMP and github.workspace to GITHUB_WORKSPACE

If a corresponding environment variable is not automatically available, you must set it in the env block of the job or step where it is needed before you can use it.

excessive-permissions

Before

name: Github CI

on:
  pull_request:
  push:
    tags:
      - "*"
    branches:
      - main

env:
  MAIN_PYTHON_VERSION: '3.12'
  DOCUMENTATION_CNAME: 'actions.docs.ansys.com'

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

jobs:
  doc-build:
    name: "Doc build"
    runs-on: ubuntu-latest
    steps:
      - uses: ansys/actions/doc-build@v10.1.0a0
        with:
          skip-install: true
          python-version: ${{ env.MAIN_PYTHON_VERSION }}
          use-python-cache: false
          needs-quarto: true

  doc-deploy-dev:
    name: "Deploy development documentation"
    runs-on: ubuntu-latest
    needs: [doc-build]
    steps:
      - uses: ansys/actions/doc-deploy-dev@v10.1.0a0
        with:
          cname: ${{ env.DOCUMENTATION_CNAME }}
          token: ${{ secrets.GITHUB_TOKEN }}
          bot-user: ${{ secrets.PYANSYS_CI_BOT_USERNAME }}
          bot-email: ${{ secrets.PYANSYS_CI_BOT_EMAIL }}

After

name: Github CI

on:
  pull_request:
  push:
    tags:
      - "*"
    branches:
      - main

env:
  MAIN_PYTHON_VERSION: '3.12'
  DOCUMENTATION_CNAME: 'actions.docs.ansys.com'

permissions: {}

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

jobs:
  doc-build:
    name: "Doc build"
    runs-on: ubuntu-latest
    steps:
      - uses: ansys/actions/doc-build@v10.1.0a0
        with:
          skip-install: true
          python-version: ${{ env.MAIN_PYTHON_VERSION }}
          use-python-cache: false
          needs-quarto: true

  doc-deploy-dev:
    name: "Deploy development documentation"
    runs-on: ubuntu-latest
    needs: [doc-build]
    permissions:
      contents: write
    steps:
      - uses: ansys/actions/doc-deploy-dev@v10.1.0a0
        with:
          cname: ${{ env.DOCUMENTATION_CNAME }}
          token: ${{ secrets.GITHUB_TOKEN }}
          bot-user: ${{ secrets.PYANSYS_CI_BOT_USERNAME }}
          bot-email: ${{ secrets.PYANSYS_CI_BOT_EMAIL }}

anonymous-definition

Before

on: push

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - run: echo "Hello!"

After

name: Echo Test
on: push

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - run: echo "Hello!"

Ignoring zizmor findings

One-off comments

While auditing workflows with zizmor, you might need to ignore findings that are not relevant to your workflows. You can ignore zizmor audits by adding a YAML comment on any line within the span of the finding.

Use the following format:

# zizmor: ignore[<rule-name>]

For example, to ignore the artipacked rule:

# zizmor: ignore[artipacked]

To ignore multiple rules in the same span, separate them with commas:

# zizmor: ignore[github-env,template-injection]

For more information, see ignoring zizmor results.

zizmor.yml configuration file

If you need to ignore multiple findings or entire files, a zizmor.yml configuration file is easier to maintain than one-off comments.

A zizmor.yml file might look like this:

rules:
  unpinned-uses:
    config:
      policies:
        ansys/*: ref-pin
        actions/*: hash-pin
  template-injection:
    ignore:
      - safe.yml
      - somewhat-safe.yml:123
      - one-exact-spot.yml:123:456

This configuration file achieves the following:

• Declares that ansys/actions can be pinned with tags, but actions/* must be pinned with a SHA.

• Ignores all template-injection findings in safe.yml, regardless of line or column location.

• Ignores any template-injection findings in somewhat-safe.yml that occur on line 123.

• Ignores one template-injection finding in one-exact-spot.yml that occurs on line 123, column 456.

For more information, see ignoring zizmor results.