Skip to content

Latest commit

 

History

History
253 lines (185 loc) · 11.2 KB

README.md

File metadata and controls

253 lines (185 loc) · 11.2 KB
Raw

IMMUNIZE

patch
logo

Pipeline for patching vulnerable container images 💉📦

Abstract

The present is a repository containing a Github action to patch
vulnerable container images with copacetic and attest them with cosign.

Note

The patched images can be found here.

Instructions

The pipeline is triggered upon a push to the repo (any branch).
The corresponding action is configured in the .github/workflows/patch.yaml file.
Specifically, the list of container images to patch is specified within the strategy as follows:

images: ['docker.io/library/nginx:1.21.6', 'docker.io/openpolicyagent/opa:0.46.0']

Following is an high-level description of the pipeline jobs and steps:

Immunize Job

Overview:

This job is triggered on every push event (excluding changes to README.md) and focuses on scanning and immunizing Docker images for security vulnerabilities.

Steps:

  1. Install Cosign:
    • Install Cosign on the runner environment
  2. Set up Docker Buildx:
    • Uses the docker/setup-buildx-action to set up Docker Buildx for multi-platform builds.
  3. Generate Trivy Report:
    • Utilizes the aquasecurity/trivy-action to scan specified Docker images for OS vulnerabilities and generates a JSON report.
  4. Check Vuln Count:
    • Parses the Trivy report using jq to count the number of vulnerabilities and outputs the count to the GitHub environment.
  5. Set Tag:
    • Extracts the tag from the Docker image reference and appends "-immunized" to create a new tag. Sets this new tag in the GitHub environment.
  6. Copa Action:
    • Conditionally executes the project-copacetic/copa-action if vulnerabilities are found.
    • Utilizes Copa to apply security patches to the Docker image, generating a patched image and a detailed report.
  7. Log into ghcr:
    • Logs into GitHub Container Registry (ghcr.io) using the docker/login-action with the GitHub token.
  8. Tag Image for GHCR:
    • Tags the patched Docker image and prepares it for pushing to GitHub Container Registry.
  9. Docker Push Patched Image:
    • Pushes the patched Docker image to GitHub Container Registry for storage and distribution.
  10. Produce Image SBOM:
    • Produce a Software Bill of Material using the anchore/sbom-action for the pushed image.
  11. Sign image with Cosign:
    • Sign the pushed image with Cosign
  12. Attest the Image with SBOM:
    • Attest the image using the SBOM as a predicate via cosign.

Send-Mail-Report Job

Overview:

This job is dependent on the completion of the Immunize job and is responsible for sending an email report.
If you dont need this job you can comment it out in the pipeline manifest.

Steps:

  1. Checkout Repository:
    • Checks out the repository to access necessary files and scripts.
  2. Send Mail Report:
    • Executes a Python script (send_mail_report.py) located in the repository, sending a report via email.
    • Configures email recipient addresses, sender address, and password using github action secrets.
  3. Report Example:
    • report

To perform image pulls, authentication is not required; however, GitHub may prompt for a token if the API call limit is exceeded.
In such instances, please refer to the instructions provided here to configure an access token.
Subsequently, proceed to log in as follows:

export CR_PAT=YOUR_TOKEN \
&& echo $CR_PAT | docker login ghcr.io -u USERNAME --password-stdin

Login Succeeded

Verify Patching

Warning

Please be aware that Copacetic focuses on rectifying vulnerabilities within the operating system's libraries in the corresponding image layer, rather than addressing application dependencies.

To assess the effectiveness of patching, you may conduct a scan using Trivy initially on one of the original images:

trivy image docker.io/openpolicyagent/opa:0.46.0

Output for OS CVEs:

Total: 41 (UNKNOWN: 0, LOW: 11, MEDIUM: 21, HIGH: 9, CRITICAL: 0)

And then on the immunized version of that same image:

trivy image ghcr.io/r3drun3/immunize/docker.io/openpolicyagent/opa:0.46.0-immunized

Output for OS CVEs:

Total: 18 (UNKNOWN: 0, LOW: 11, MEDIUM: 7, HIGH: 0, CRITICAL: 0)

As you can see the latest has way less CVEs than the former!

The benefit becomes even more apparent when patching old images.
For instance, observe the difference between this:

trivy image python:3.5.10-slim

Total: 307 (UNKNOWN: 8, LOW: 101, MEDIUM: 81, HIGH: 87, CRITICAL: 30)

and the immunized one:

trivy image ghcr.io/r3drun3/immunize/docker.io/library/python:3.5.10-slim-immunized

Total: 148 (UNKNOWN: 0, LOW: 97, MEDIUM: 26, HIGH: 23, CRITICAL: 2)

Critical vulnerabilities have been reduced by ~93%!

Warning

Even though the benefits are more apparent in older images, it is always advisable to prioritize new images and the most recent tags whenever possible.

Verify Image Signatures and Attestations

All the patched OCI images produced by the pipeline are signed with cosign.
In order to verify the signature, adapt the following command for the desired image:

cosign verify --key cosign/cosign.pub ghcr.io/r3drun3/immunize/docker.io/library/node:18.17.1-slim-immunized

Output:

Verification for ghcr.io/r3drun3/immunize/docker.io/library/node:18.17.1-slim-immunized --
The following checks were performed on each of these signatures:
  - The cosign claims were validated
  - Existence of the claims in the transparency log was verified offline
  - The signatures were verified against the specified public key

[
   {"critical":
      {"identity":{"docker-reference":"ghcr.io/r3drun3/immunize/docker.io/library/node"},
      "image":
      {"docker-manifest-digest":"sha256:19940c59087a363148b44c56447186d97d6afbc2165727b2d0a2ea0ce43b69fd"},
      "type":"cosign container image signature"},
      "optional":
         {"Bundle":
         {"SignedEntryTimestamp":"MEUCIHo1Jja4t0+OPDYqHo/B/p7HUtP+/i8ZD+fu6Rb57Lw9AiEA7N1i7JDiIvRxu9QtYOrrS8Y+AeekHMWNE3p7GJAbHAs=",
         "Payload":     {"body":"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",
   "integratedTime":1705317423,
   "logIndex":63825695,
   "logID":"c0d23d6ad406973f9559f3ba2d1ca01f84147d8ffc5b8445c224f98b9591801d"}
     }
    }
   }
]

The pipeline, for every image, produces an SBOM in SPDX format and creates an in-toto attestation for the image using that artifact as a predicate.

Note

You can learn the difference between SBOMs and Attestations here.

In order to verify the image attestation with cosign, use the following command (adapt for the desired image):

export IMAGE=ghcr.io/r3drun3/immunize/docker.io/openpolicyagent/opa:0.46.0-immunized
cosign verify-attestation --type spdx --key ./cosign/cosign.pub $IMAGE

The above command verifies and returns the uploaded artifact data in base64 format.
We can decode it to query the artifact (in this case, the SBOM file):

cosign verify-attestation --type spdx --key ./cosign/cosign.pub $IMAGE | jq -r .payload | base64 -D | jq .

The SPDX sbom is the predicate property in the previous command output:

Output Sample:

Verification for ghcr.io/r3drun3/immunize/docker.io/openpolicyagent/opa:0.46.0-immunized --
The following checks were performed on each of these signatures:
  - The cosign claims were validated
  - Existence of the claims in the transparency log was verified offline
  - The signatures were verified against the specified public key
{
  "_type": "https://in-toto.io/Statement/v0.1",
  "predicateType": "https://spdx.dev/Document",
  "subject": [
    {
      "name": "ghcr.io/r3drun3/immunize/docker.io/openpolicyagent/opa",
      "digest": {
        "sha256": "45de0d4de2ef8590ccdb97ee817d70ba6bb3efb70aabd4abe74a3a6facff24ea"
      }
    }
  ],
  "predicate": "[SPDX-SBOM-HERE]"
}

In order to better understand the previous output, take a look at the in-toto attestation framework spec.
In ordert to understand how in-toto verifies an attestation, take a look at the in-toto validation model.
If you want to learn how to enforce the SBOM attestations on your kubernetes cluster take a look at this repo.

If you want to produce a single json file for the sbom in SPDX format, you can use the following command:

cosign verify-attestation --type spdx --key ./cosign/cosign.pub $IMAGE | jq -r .payload | base64 -D | jq -r '.predicate | fromjson' > final-spdx.json

Starting from that command you can then use open source tools or this website to do validation and other operations on the SBOM.

For example, let's say I want to validate the SBOM for the ghcr.io/r3drun3/immunize/docker.io/library/python:3.5.10-slim-immunized image.
I will start by obtaining the SBOM in Json format:

export IMAGE=ghcr.io/r3drun3/immunize/docker.io/library/python:3.5.10-slim-immunized
cosign verify-attestation --type spdx --key ./cosign/cosign.pub $IMAGE | jq -r .payload | base64 -D | jq -r '.predicate | fromjson' > final-spdx.json

Then I will head to the aforementioned website, select validate and upload my json file:
valid

If I want to produce an XLSX spreadsheet, I will select convert on the website, enter the format, upload the file and then download the produced spreadsheet:
valid