The out_http output plugin allows the use of placeholders (such as ${tag}) in the endpoint configuration parameter.
It was discovered that if the placeholder value is derived from untrusted user input, an attacker can maliciously control the destination hostname of the outbound HTTP requests made by Fluentd.
Impact
This vulnerability allows for a Server-Side Request Forgery (SSRF) attack.
An unauthenticated attacker can force the Fluentd node to send HTTP requests to arbitrary internal services. This can lead to unauthorized access to internal APIs, data exfiltration, or the compromise of cloud metadata endpoints (e.g., AWS IMDS 169.254.169.254).
Patches
v1.19.3
Workarounds
If an immediate upgrade is not possible, users are strongly advised to apply the following mitigations:
-
Avoid Dynamic Hostnames
- Do not use the placeholder in the endpoint parameter as hostname.
-
Restrict Network Access
- Use firewall rules (e.g., iptables, AWS Security Groups) to block the Fluentd node from accessing sensitive internal IP addresses, specifically the cloud provider's metadata service and other internal microservices that Fluentd does not explicitly need to access.
-
Restrict allowed hosts
- Inject filter to accept allowed hosts in placeholders explicitly if possible.
Note regarding Path/Query manipulation via Placeholders
The security fix implemented for this advisory specifically validates the host component of the resolved URL to prevent redirection to arbitrary external servers.
Please be aware that Fluentd intentionally does not strictly sanitize or URI-encode URL metacharacters (such as ?, #, or /) within placeholders like ${tag}.
This is a design decision to preserve flexibility for users who legitimately rely on dynamic path construction or query parameters for internal routing.
While this means a manipulated tag could theoretically alter the requested path or query string on the same validated host, it does not constitute a valid SSRF vulnerability in this context for the following reasons:
- HTTP Method Constraint
- The
out_http plugin exclusively issues POST and PUT requests. It does not issue GET requests, rendering standard cloud metadata extraction attacks (e.g., against AWS IMDS or GCP Metadata) unfeasible, as those endpoints strictly require GET requests.
- Intended Design
- Altering the path/query within the confines of the trusted target host is considered expected behavior when utilizing dynamic placeholders.
Users who incorporate ${tag} or other placeholders into their endpoint configuration should ensure they trust the source of their incoming tags, or use <filter> plugins to sanitize tags before routing if strict path enforcement is required in their specific environment.
The
out_httpoutput plugin allows the use of placeholders (such as${tag}) in theendpointconfiguration parameter.It was discovered that if the placeholder value is derived from untrusted user input, an attacker can maliciously control the destination hostname of the outbound HTTP requests made by Fluentd.
Impact
This vulnerability allows for a Server-Side Request Forgery (SSRF) attack.
An unauthenticated attacker can force the Fluentd node to send HTTP requests to arbitrary internal services. This can lead to unauthorized access to internal APIs, data exfiltration, or the compromise of cloud metadata endpoints (e.g., AWS IMDS
169.254.169.254).Patches
v1.19.3
Workarounds
If an immediate upgrade is not possible, users are strongly advised to apply the following mitigations:
Avoid Dynamic Hostnames
Restrict Network Access
Restrict allowed hosts
Note regarding Path/Query manipulation via Placeholders
The security fix implemented for this advisory specifically validates the host component of the resolved URL to prevent redirection to arbitrary external servers.
Please be aware that Fluentd intentionally does not strictly sanitize or URI-encode URL metacharacters (such as
?,#, or/) within placeholders like${tag}.This is a design decision to preserve flexibility for users who legitimately rely on dynamic path construction or query parameters for internal routing.
While this means a manipulated tag could theoretically alter the requested path or query string on the same validated host, it does not constitute a valid SSRF vulnerability in this context for the following reasons:
out_httpplugin exclusively issuesPOSTandPUTrequests. It does not issueGETrequests, rendering standard cloud metadata extraction attacks (e.g., against AWS IMDS or GCP Metadata) unfeasible, as those endpoints strictly requireGETrequests.Users who incorporate
${tag}or other placeholders into theirendpointconfiguration should ensure they trust the source of their incoming tags, or use<filter>plugins to sanitize tags before routing if strict path enforcement is required in their specific environment.