Docker containers have been an essential part of the developer's toolbox for several years now, allowing them to build, distribute and deploy their applications in a standardized way.
This gain in traction has been, not surprisingly, accompanied by a surge in security issues related to containerization technology. Indeed, containers also represent a standardized surface for attackers. They can easily exploit misconfigurations and escape from within containers to the host machine.
Furthermore, the word “container” is often misunderstood, as many developers tend to associate the concept of isolation with a false sense of security, believing that this technology is inherently safe.
The key here is that containers don’t have any security dimension by default. Their security completely depends on:
- the supporting infrastructure (OS and platform)
- their embedded software components
- their runtime configuration
Container security represents a broad topic, but the good news is that many best practices are low-hanging fruits one can harvest to quickly reduce the attack surface of their deployments.
That's why we curated a set of the best recommendations regarding Docker containers configuration at build and runtime. Check out the one-page cheat sheet below.Download the Docker security cheatsheet
Note: in a managed environment like Kubernetes, most of these settings can be overridden by a Security Context or other higher-level security rules. See more
1. Build Configuration
1.1 Use trusted images
Carefully choose your base image when you
docker pull image:tag
You should always prefer using a trusted image, preferably from the Docker Official Images, in order to mitigate supply chain attacks.
If you need to choose a base distro, Alpine Linux is recommended since it is one of the lightest available, ensuring the attack surface is reduced.
Do I need to use the latest or a fixed tag version?
First, you should understand that Docker tagging works from less to more specific, that’s the reason why :
python:3.9.6-alpine3.14 python:3.9.6-alpine python:3.9-alpine python:alpine
all refer to the same image (at the moment of writing) for example.
By being very specific and pinning down a version, you are shielding yourself from any future breaking change. On the other hand, using the latest version ensures that more vulnerabilities are patched. This is a tradeoff, but pinning to a stable release is what is generally recommended.
Considering that, we would pick
Note: the same applies to packages installed during the build process of your image.
1.2 Always use an unprivileged user
By default, the process inside a container is run as root (id=0).
To enforce the principle of least privilege, you should set a default user. For this you have two options:
- Either specify an arbitrary user ID that won’t exist in the running container, with the
docker run -u 4000 <image>
Note: if you later need to mount a filesystem, you should match the user ID you are using to the host user in order to access the files.
- Or anticipate by creating a default user in your Dockerfile:
FROM <base image> RUN addgroup -S appgroup \ && adduser -S appuser -G appgroup USER appuser ... <rest of Dockerfile> ...
Note: you would need to check what tool is used to create users and groups in your base image.
1.3 Use a separate User ID namespace
By default, the Docker daemon uses the host’s user ID namespace. Consequently, any success in privilege escalation inside a container would also mean root access both to the host and to other containers.
To mitigate this risk, you should configure your host and the Docker daemon to use a separate namespace with the
--userns-remap option. See more
1.4 Handle environment variables with care
You should never include sensitive information in plaintext in an ENV directive: they are simply not a safe place to store any bit of information you don’t want to be present in the final layer. For example, if you thought that unsetting an environment variable like this:
ENV $VAR RUN unset $VAR
Was safe, you are wrong!
$VAR will still be present in the containers and could be dumped anytime!
To prevent runtime read access, use a single RUN command to set and unset the variable in a single layer (don't forget the variable can still be extracted from the image).
RUN export ADMIN_USER="admin" \ && ... \ && unset ADMIN_USER
More idiomatically, use the ARG directive (ARG values are not available after the image is built).
Unfortunately, secrets are too often hardcoded into docker images’ layers, that’s the reason we developed a scanning tool leveraging GitGuardian secrets engine to find them:
ggshield scan docker <image>
More on scanning images for vulnerabilities later.
1.5 Don’t expose the Docker daemon socket
Unless you are very confident with what you are doing, never expose the UNIX socket that Docker is listening to:
This is the primary entry point for the Docker API. Giving someone access to it is equivalent to giving unrestricted root access to your host.
You should never expose it to other containers:
2. Privileges, capabilities, and shared resources
2.1 Forbid new privileges
First, your container should never be running as privileged, otherwise, it would be allowed to have all the root capabilities on the host machine.
To be even safer, it is recommended to explicitly forbid the possibility to add new privileges after a container has been created with the option
2.2 Define fine-grained capabilities
Second, capabilities are a Linux mechanism used by Docker to turn the binary “root/non-root” dichotomy into a fine-grained access control system: your containers are run with a default set of enabled capabilities, which you most probably don’t all need.
2.3 Drop all default capabilities
It's recommended to drop all default capabilities and only add them individually: see the list of default capabilities
for instance, a web server would probably only need the NET_BIND_SERVICE to bind to a port under 1024 (like port 80).
2.4 Avoid sharing sensitive filesystem parts
Third, don’t share the sensitive parts of the host filesystem :
- root (/),
- device (/dev)
- process (/proc)
- virtual (/sys) mount points.
If you need access to host devices, be careful to selectively enable the access options with the
[r|w|m]flags (read, write, and use mknod).
2.5 Use Control Groups to limit access to resources
Control Groups are the mechanism used to control access to CPU, memory, and disk I/O for each container.
By default, a container is associated with a dedicated
cgroup, but if the option
--cgroup-parent is present, you are putting the host resources at risk of a DoS attack, because you are allowing shared resources between the host and the container.
In the same idea, it is recommended to specify memory and CPU usage by using options like
--memory=”400m” --memory-swap=”1g” --cpus=0.5 --restart=on-failure:5 --ulimit nofile=5 --ulimit nproc=5
See more on resources constraints
3.1 Only allow read access to the root filesystem
Containers should be ephemeral and thus mostly stateless. That’s why you can often limit the mounted filesystem to be read-only.
docker run --read-only <image>
3.2 Use a temporary filesystem for non-persistent data
If you need only temporary storage, use the appropriate option
docker run --read-only --tmpfs /tmp:rw ,noexec,nosuid <image>
3.3 Use a filesystem for persistent data
If you need to share data with the host filesystem or other containers, you have two options:
- Create a bind mount with limited useable disk space (
- Create a bind volume for a dedicated partition (
In either case, if the shared data doesn’t need to be modified by the container, use the read-only option.
docker run -v <volume-name>:/path/in/container:ro <image>
docker run --mount source=<volume-name>,destination=/path/in/container,readonly <image>
4. Network Security
4.1 Don’t use Docker’s default bridge docker0
Docker container networking security requires taking a step back to understand what is done at launch.
By default, Docker creates a
docker0 network bridge to separate the host network from the container network.
When a container is created, Docker connects it to the
docker0 network by default. Therefore, all containers are connected to
docker0 and are able to communicate with each other.
Now, a basic network security measure is to disable this default connection of all the containers by specifying the option
--bridge=none. Instead, you should create a dedicated network for every connection with the command:
docker network create <network_name>
And then use it to access the host network interface
docker run --network=<network_name>
For example, to create a web server talking to a database (started in another container), the best practice would be to create a bridge network
WEB in order to route incoming traffic from the host network interface and use another bridge
DB only used to connect the database and the web containers.
4.2 Don’t share the host’s network namespace
Same idea, isolate the host's network interface: the
--network host option should not be used.
5.1 Export logs
The default logging level is INFO, but you can specify another one with the option:
What is less known is the log export capacity of Docker: if your containerized app produces event logs, you can redirect
STDOUT streams to an external logging service for decoupling using the option
5.2 Setup dual logging
You can also enable dual logging to preserve docker access to logs while using an external service. If your app uses dedicated files (often written under
/var/log), you can still redirect these streams: see the official documentation
6. Scan for vulnerabilities & secrets
Last but not least, I hope it is now clear that your containers are only going to be as safe as the software they are running. To make sure your images are vulnerability-free, you need to perform a scan for known vulnerabilities.
Many tools are available for different use-case and in different forms:
6.1 Scan for vulnerabilities
- Snyk (open source and free option available)
- Anchore (open source and free option available)
- JFrog XRay
6.2 Scan for secrets
Docker images are full of hard-coded plaintext secrets. Even public images distributed by reputable cloud vendors can leak secrets through their build process.
A study we conducted in 2021 on Docker Hub showed that more than 4,000 secrets (1,200+ unique) were hard-coded in a 10,000 image sample. Also, 4.62% of the Docker Hub images sample exposed one or more secrets:
It is clear today that scanning for secrets in images (whether home-built or third-party-provided) is essential to guard against supply chain attacks.
Learn how to find leaked credentials in Docker images with ggshield (free for individual developers and small teams.):
Here is our final Docker security checklist:
1. Build Configuration
- Use trusted images
- Always use an unprivileged user
- Use a separate User ID namespace
- Handle environment variables with care
- Don’t expose the Docker daemon socket
2. Privileges, capabilities, and shared resources
- Forbid new privileges
- Define fine-grained capabilities
- Drop all default capabilities
- Avoid sharing sensitive filesystem paths
- Use Control Groups to limit access to resources
- Only allow read access to the root filesystem
- Use a temporary filesystem for non-persistent data
- Use a filesystem for persistent data
4. Network Security
- Don’t use Docker’s default bridge docker0
- Don’t share the host’s network namespace
- Export logs
- Setup dual logging
6. Scan for vulnerabilities & secrets
- Scan for vulnerabilities
- Scan for secrets
Is Docker good for security?Docker can be a useful tool for security if used correctly, but it's not a security solution in and of itself. Docker's containerization technology can help to isolate and contain potentially vulnerable applications, but it's important to remember that Docker images can still contain security vulnerabilities or be misconfigured. Here are some ways Docker can help with security:
1. Isolation: Docker containers provide a level of isolation between applications running on the same server or host. This means that if one container is compromised, it's less likely that the attack will spread to other containers or the host itself.
2. Portability: Docker images can be easily moved between environments, making it easier to deploy applications consistently and securely.
3. Versioning: Docker images can be versioned and tracked, making it easier to roll back to a known-good state if a security issue is discovered.
4. Resource limits: Docker allows you to set resource limits for each container, preventing one container from consuming all available resources and affecting other containers or the host itself.
5. Image scanning: Docker images can be scanned for known vulnerabilities before deployment, helping to prevent the deployment of potentially vulnerable applications.
It's important to note, however, that Docker is just one part of a comprehensive security strategy. Other security measures, such as network security, access controls, and monitoring, should also be implemented to fully protect your applications and infrastructure.
How do I ensure Docker security?If you are looking for the best practices regarding Docker security, you are in the right place! Here is a helpful summary of the security measures we mentionned:
1. Use official images: Use official images from Docker Hub or other trusted repositories rather than third-party images.
2. Keep your images up to date: Regularly update your Docker images to ensure that they have the latest security patches.
3. Use strong passwords: Use strong and unique passwords for all services running inside the container.
4. Limit container privileges: Limit container privileges by running containers as non-root users.
5. Avoid running unnecessary services: Avoid running unnecessary services inside the container, as this can create security vulnerabilities.
6. Use network segmentation: Use network segmentation to separate your containers from the host system and other containers.
7. Enable Docker Content Trust: Enable Docker Content Trust to ensure the integrity of your images.
8. Use a firewall: Use a firewall to restrict network access to your containers.
9. Monitor container activity: Monitor container activity for signs of suspicious behavior.
10. Regularly audit container configurations: Regularly audit container configurations to ensure they adhere to security best practices.
What is the best practice for handling secrets in Docker?Handling secrets in Docker involves securely managing sensitive information such as passwords, API keys, and other credentials that are required by applications running inside Docker containers. Here are some best practices for handling secrets in Docker:
1. Use environment variables: One way to handle secrets is to use environment variables to store sensitive information. You can pass the values of these variables to your Docker containers at runtime, without exposing them in your Dockerfile or on your host system.
2. Use Docker secrets: Docker provides a built-in secrets management system that allows you to securely store and manage sensitive information. You can use the Docker CLI or Docker Compose to create and manage secrets, and then pass them to your containers at runtime.
3. Use third-party secrets management tools: There are also third-party secrets management tools that you can use to securely store and manage sensitive information. These tools typically provide features such as encryption, access control, and audit logging.
4. Keep secrets out of source control: To avoid exposing secrets to unauthorized users, never store them in source control systems such as Git.
5. Use multi-stage builds: When building Docker images, use multi-stage builds to separate build-time dependencies from runtime dependencies. This helps reduce the risk of exposing secrets during the build process.
6. Implement access controls: Implement access controls to restrict access to secrets only to authorized users.
Before you go
Whether you want to deep dive into a powerful container security mechanism (Seccomp filters), or learn the secrets management best practices with Docker, we have you covered:
And finally, here you'll find other cheat sheets as well:
- Best practices for managing and storing secrets including API keys and other credentials
- How to rewrite your git history
- How to safely setup multiple GitHub accounts on your local machine