Skip to content

Commit

Permalink
Merge pull request #739 from dev-sec/restruct_vars
Browse files Browse the repository at this point in the history
restructure readme to move known problems up top
  • Loading branch information
schurzi authored Jan 15, 2024
2 parents aca4bc7 + e5cc9bb commit 94a49a4
Show file tree
Hide file tree
Showing 2 changed files with 117 additions and 115 deletions.
177 changes: 90 additions & 87 deletions roles/os_hardening/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,96 @@ We disable the following filesystems, because they're most likely not used:

To prevent some of the filesystems from being disabled, add them to the `os_filesystem_whitelist` variable.

## Known Limitations

### Docker support

If you're using Docker / Kubernetes+Docker you'll need to override the ipv4 ip forward sysctl setting.

```yaml
- hosts: localhost
roles:
- devsec.hardening.os_hardening
vars:
sysctl_overwrite:
# Enable IPv4 traffic forwarding.
net.ipv4.ip_forward: 1
```
### hidepid on RHEL/CentOS 7
When having `polkit-0.112-18.el7` (and later) installed and `/proc` mounted with `hidepid=2`, every time someone uses `systemctl` the following error is displayed, but systemctl runs successfully.

```
Error registering authentication agent: GDBus.Error:org.freedesktop.PolicyKit1.Error.Failed: Cannot determine user of subject (polkit-error-quark, 0)
```
We decided to set `hidepid=0` to remove the error message, if you want to use the other proposed workaround, you have to setup the user yourself and set our option to `hidepid_option: 2` via Ansible vars.
For further details see [RedHat: "GDBus.Error:org.freedesktop.PolicyKit1.Error.Failed: Cannot determine user of subject" seen when executing systemctl command](https://access.redhat.com/solutions/5005111) or [#364: hidepid=2 gives error when running systemctl on EL7](https://github.com/dev-sec/ansible-collection-hardening/issues/364)
### sysctl - vm.mmap_rnd_bits
We are setting this sysctl to a default of `32`, some systems only support smaller values and this will generate an error. Unfortunately we cannot determine the correct applicable maximum. If you encounter this error you have to override this sysctl in your playbook.
```yaml
- hosts: localhost
roles:
- devsec.hardening.os_hardening
vars:
sysctl_overwrite:
vm.mmap_rnd_bits: 16
```

## Testing with inspec

If you're using inspec to test your machines after applying this role, please make sure to add the connecting user to the `os_ignore_users`-variable.
Otherwise inspec will fail. For more information, see [issue #124](https://github.com/dev-sec/ansible-os-hardening/issues/124).

We know that this is the case on Raspberry Pi.

## Changing sysctl variables

If you want to override sysctl-variables, you can use the `sysctl_overwrite` variable (in older versions you had to override the whole `sysctl_dict`).
So for example if you want to change the IPv4 traffic forwarding variable to `1`, do it like this:

```yaml
- hosts: localhost
roles:
- devsec.hardening.os_hardening
vars:
sysctl_overwrite:
# Enable IPv4 traffic forwarding.
net.ipv4.ip_forward: 1
```
Alternatively you can change Ansible's [hash-behaviour](https://docs.ansible.com/ansible/latest/reference_appendices/config.html#default-hash-behaviour) to `merge`, then you only have to overwrite the single hash you need to. But please be aware that changing the hash-behaviour changes it for all your playbooks and is not recommended by Ansible.

## Improving Kernel Audit logging

By default, any process that starts before the `auditd` daemon will have an AUID of `4294967295`. To improve this and provide more accurate logging, it's recommended to add the kernel boot parameter `audit=1` to you configuration. Without doing this, you will find that your `auditd` logs fail to properly audit all processes.

For more information, please see this [upstream documentation](https://www.kernel.org/doc/html/latest/admin-guide/kernel-parameters.html) and your system's boot loader documentation for how to configure additional kernel parameters.

## PAM configuration

We use specific PAM configuration for every linux distribution. This is a very diverse area, since every vendor seems to push their own system for managing PAM. We used following guides and sources for our configuration.

- on Debian/Ubuntu we use [pam-config-framework](https://wiki.ubuntu.com/PAMConfigFrameworkSpec)
- RHEL8 and compatible use information from [authselect](https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/configuring_authentication_and_authorization_in_rhel/configuring-user-authentication-using-authselect_configuring-authentication-and-authorization-in-rhel), but our setup disables the configuration from authselect
- RHEL 6/7 and compatible use [authconfig](https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/system-level_authentication_guide/authconfig-install), but our setup disables the configuration from authconfig
- for RHEL 6/7 we also use information from [Desktop Security](https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/security_guide/chap-hardening_your_system_with_tools_and_services#sec-Desktop_Security)


## Acknowledgements

This role is mostly based on guides by:

- [Arch Linux wiki, Sysctl hardening](https://wiki.archlinux.org/index.php/Sysctl)
- [NSA: Guide to the Secure Configuration of Red Hat Enterprise Linux 5](http://www.nsa.gov/ia/_files/os/redhat/rhel5-guide-i731.pdf)
- [Ubuntu Security/Features](https://wiki.ubuntu.com/Security/Features)
- [Deutsche Telekom, Group IT Security, Security Requirements (German)](https://www.telekom.com/psa)

<!-- BEGIN_ANSIBLE_DOCS -->

## Supported Operating Systems
Expand Down Expand Up @@ -924,90 +1014,3 @@ None.
```
<!-- END_ANSIBLE_DOCS -->

## Known Limitations

### Docker support

If you're using Docker / Kubernetes+Docker you'll need to override the ipv4 ip forward sysctl setting.

```yaml
- hosts: localhost
roles:
- devsec.hardening.os_hardening
vars:
sysctl_overwrite:
# Enable IPv4 traffic forwarding.
net.ipv4.ip_forward: 1
```
### hidepid on RHEL/CentOS 7
When having `polkit-0.112-18.el7` (and later) installed and `/proc` mounted with `hidepid=2`, every time someone uses `systemctl` the following error is displayed, but systemctl runs successfully.

```
Error registering authentication agent: GDBus.Error:org.freedesktop.PolicyKit1.Error.Failed: Cannot determine user of subject (polkit-error-quark, 0)
```
We decided to set `hidepid=0` to remove the error message, if you want to use the other proposed workaround, you have to setup the user yourself and set our option to `hidepid_option: 2` via Ansible vars.
For further details see [RedHat: "GDBus.Error:org.freedesktop.PolicyKit1.Error.Failed: Cannot determine user of subject" seen when executing systemctl command](https://access.redhat.com/solutions/5005111) or [#364: hidepid=2 gives error when running systemctl on EL7](https://github.com/dev-sec/ansible-collection-hardening/issues/364)
### sysctl - vm.mmap_rnd_bits
We are setting this sysctl to a default of `32`, some systems only support smaller values and this will generate an error. Unfortunately we cannot determine the correct applicable maximum. If you encounter this error you have to override this sysctl in your playbook.
```yaml
- hosts: localhost
roles:
- devsec.hardening.os_hardening
vars:
sysctl_overwrite:
vm.mmap_rnd_bits: 16
```

## Testing with inspec

If you're using inspec to test your machines after applying this role, please make sure to add the connecting user to the `os_ignore_users`-variable.
Otherwise inspec will fail. For more information, see [issue #124](https://github.com/dev-sec/ansible-os-hardening/issues/124).

We know that this is the case on Raspberry Pi.

## Changing sysctl variables

If you want to override sysctl-variables, you can use the `sysctl_overwrite` variable (in older versions you had to override the whole `sysctl_dict`).
So for example if you want to change the IPv4 traffic forwarding variable to `1`, do it like this:

```yaml
- hosts: localhost
roles:
- devsec.hardening.os_hardening
vars:
sysctl_overwrite:
# Enable IPv4 traffic forwarding.
net.ipv4.ip_forward: 1
```
Alternatively you can change Ansible's [hash-behaviour](https://docs.ansible.com/ansible/latest/reference_appendices/config.html#default-hash-behaviour) to `merge`, then you only have to overwrite the single hash you need to. But please be aware that changing the hash-behaviour changes it for all your playbooks and is not recommended by Ansible.

## Improving Kernel Audit logging

By default, any process that starts before the `auditd` daemon will have an AUID of `4294967295`. To improve this and provide more accurate logging, it's recommended to add the kernel boot parameter `audit=1` to you configuration. Without doing this, you will find that your `auditd` logs fail to properly audit all processes.

For more information, please see this [upstream documentation](https://www.kernel.org/doc/html/latest/admin-guide/kernel-parameters.html) and your system's boot loader documentation for how to configure additional kernel parameters.

## More information

We use specific PAM configuration for every linux distribution. This is a very diverse area, since every vendor seems to push their own system for managing PAM. We used following guides and sources for our configuration.

- on Debian/Ubuntu we use [pam-config-framework](https://wiki.ubuntu.com/PAMConfigFrameworkSpec)
- RHEL8 and compatible use information from [authselect](https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/configuring_authentication_and_authorization_in_rhel/configuring-user-authentication-using-authselect_configuring-authentication-and-authorization-in-rhel), but our setup disables the configuration from authselect
- RHEL 6/7 and compatible use [authconfig](https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/system-level_authentication_guide/authconfig-install), but our setup disables the configuration from authconfig
- for RHEL 6/7 we also use information from [Desktop Security](https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/security_guide/chap-hardening_your_system_with_tools_and_services#sec-Desktop_Security)

This role is mostly based on guides by:

- [Arch Linux wiki, Sysctl hardening](https://wiki.archlinux.org/index.php/Sysctl)
- [NSA: Guide to the Secure Configuration of Red Hat Enterprise Linux 5](http://www.nsa.gov/ia/_files/os/redhat/rhel5-guide-i731.pdf)
- [Ubuntu Security/Features](https://wiki.ubuntu.com/Security/Features)
- [Deutsche Telekom, Group IT Security, Security Requirements (German)](https://www.telekom.com/psa)
55 changes: 27 additions & 28 deletions roles/ssh_hardening/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,33 @@ This role provides secure ssh-client and ssh-server configurations. It is intend

Warning: This role disables root-login on the target server! Please make sure you have another user with su or sudo permissions that can login into the server.

## Configuring settings not listed in role-variables

If you want to configure ssh options that are not listed above, you can use `ssh_custom_options` (for `/etc/ssh/ssh_config`) or `sshd_custom_options` (for `/etc/ssh/sshd_config`) to set them. These options will be set on the **beginning** of the file so you can override options further down in the file.

Example playbook:

```yml
- hosts: localhost
roles:
- devsec.hardening.ssh_hardening
vars:
ssh_custom_options:
- "Include /etc/ssh/ssh_config.d/*"
sshd_custom_options:
- "AcceptEnv LANG"
```
## After using the role Ansibles template/copy/file module does not work anymore
If you use this role to disable SFTP (`sftp_enabled: false`) you will get errors. Ansible uses by default SFTP to transfer files to the remote hosts. You have to set `scp_if_ssh = True` in your ansible.cfg. This way Ansible uses SCP to copy files. If your control node uses OpenSSH version 9.0 or above, you also need to set `scp_extra_args = "-O"`, since starting with that version the `scp` utility also defaults to using SFTP.

## Changing the default port and idempotency

This role uses the default port 22 or the port configured in the inventory to connect to the server. If the default `ssh` port is changed via `ssh_server_ports`, once the ssh server is restarted, it will still try to connect using the previous port. In order to run this role again on the same server the inventory will have to be updated to use the new ssh port.

If idempotency is important, please consider using role [`ssh-hardening-fallback`](https://github.com/nununo/ansible-ssh-hardening-fallback), which is a wrapper around this role that falls back to port 22 if the configured port is unreachable.

<!-- BEGIN_ANSIBLE_DOCS -->

## Supported Operating Systems
Expand Down Expand Up @@ -442,32 +469,4 @@ None.
roles:
- name: devsec.hardening.ssh_hardening
```

<!-- END_ANSIBLE_DOCS -->

## Configuring settings not listed in role-variables

If you want to configure ssh options that are not listed above, you can use `ssh_custom_options` (for `/etc/ssh/ssh_config`) or `sshd_custom_options` (for `/etc/ssh/sshd_config`) to set them. These options will be set on the **beginning** of the file so you can override options further down in the file.

Example playbook:

```yml
- hosts: localhost
roles:
- devsec.hardening.ssh_hardening
vars:
ssh_custom_options:
- "Include /etc/ssh/ssh_config.d/*"
sshd_custom_options:
- "AcceptEnv LANG"
```
## After using the role Ansibles template/copy/file module does not work anymore
If you use this role to disable SFTP (`sftp_enabled: false`) you will get errors. Ansible uses by default SFTP to transfer files to the remote hosts. You have to set `scp_if_ssh = True` in your ansible.cfg. This way Ansible uses SCP to copy files. If your control node uses OpenSSH version 9.0 or above, you also need to set `scp_extra_args = "-O"`, since starting with that version the `scp` utility also defaults to using SFTP.

## Changing the default port and idempotency

This role uses the default port 22 or the port configured in the inventory to connect to the server. If the default `ssh` port is changed via `ssh_server_ports`, once the ssh server is restarted, it will still try to connect using the previous port. In order to run this role again on the same server the inventory will have to be updated to use the new ssh port.

If idempotency is important, please consider using role [`ssh-hardening-fallback`](https://github.com/nununo/ansible-ssh-hardening-fallback), which is a wrapper around this role that falls back to port 22 if the configured port is unreachable.

0 comments on commit 94a49a4

Please sign in to comment.