Open ferricoxide opened 7 months ago
It is likely related to the rule rsyslog_remote_loghost
.
@marcusburghardt sed:
It is likely related to the rule
rsyslog_remote_loghost
.
Yeah, when I'd generated the BASH fix-script from the remediation-content, I believe that was the title of the block in the generated-script. We don't generally use tayloring-files to prevent scans/remediations from running.
That usage-method aside, I still feel that, given the inherently-local nature of remote log-offload, having the default content inject a (generally spurious) remote-logging entry into hardened-hosts' rsyslog.conf
files doesn't make a ton of sense.
This issue is likely related: https://github.com/ComplianceAsCode/content/issues/8018 @ferricoxide , would you like to propose a PR improving this rule?
@marcusburghardt
I've mostly used this project's content in a fairly static way: is there a way to add usage of externalities to guide the setup that the … --remediate …
automation does? Otherwise, the only PR I'd submit would be one that simply avoids tacking on the offending line.
Share the context
We use this project's content across several customers' projects. Initially discovered the problem last year, but reporting the issue "fell off the stack". Recently, began receiving complaints from another customer that they were seeing "host not found" entries in their system logs for the
logcollector
host specified in the post-hardening/etc/rsyslog.conf
configuration.Description of problem:
Looks like this project's remediation-automation is injecting
@@logcollector
into the hardened-systems' /etc/rsyslog.confconfiguration. At bare minimum, this results in nuisance "host not found" error-types. Also poses a potential security-risk: if someone injects
logcollectorinto the DNS domain the hardened system is configured into, log-offload attempts could be made to a bogus/unsanctioned rsyslog. Security-concern aside, if someone injects
logcollector` into the DNS domain the hardened system is configured into, but doesn't stand up a collector, the "host not found" errors will be replaced with "cannot connect" error-typesProposed change:
Centralize log-collection is inherently site-specific. In many cases, other services (e.g., Splunk, CloudWatch Agent, etc.) are used for offload rather than syslog. As such, the automation should include no remediative-actions that include injecting a logging-destination into the hardened-hosts'
/etc/rsyslog.conf
configurations.CC: @lorengordon and @eemperor