CRITICAL: Fluent Bit Flaws Enable RCE and Telemetry Tampering in Major Orgs

Oligo Security researchers have uncovered a dangerous chain of vulnerabilities in Fluent Bit, the popular, lightweight telemetry agent used by major organizations—including in finance, delivery apps, and cloud platforms. Because Fluent Bit sits directly on the ingestion path and handles sensitive data and untrusted input, any bug can become a “high impact security issue.”

The newly disclosed flaws create a severe attack path, allowing threat actors to manipulate log routing, write malicious files to the filesystem, and even crash the agent or achieve Remote Code Execution (RCE).

The core of the attack chain exploits how Fluent Bit handles tags—the strings used to identify and classify data for routing.

1. Authentication Bypass (CVE-2025-12969)

The in_forward plugin, which receives logs from other agents, has a critical authentication bypass. If operators configure it with Security.Users (username and password authentication) but do not set a Shared_Key, the report found that “authentication is not enforced.” This “leaves many fluent-bit forwarders open to attacker connections while giving users a false sense of security.” An attacker could “flood a security product’s logs with false events to spam alerts to overwhelm security teams” or inject false data to hide their activity.

2. Tag Key Spoofing (CVE-2025-12978)

This flaw allows an attacker to control a record’s tag, bypassing configuration security. The issue lies in a partial string comparison in the tag_key() method. The comparison checks the size of the user input key, “meaning that an attacker can send just the first letter and match the Tag_Key.” In short, “by guessing just the first letter of a tag key an attacker can spoof the tag on log data and control where and how it’s processed.”

3. Tag Injection (CVE-2025-12977)

Tags derived dynamically from a record using Tag_Key bypass the normal sanitization process. This allows an attacker to inject problematic characters like spaces, newlines, or path traversal patterns (../) into the tag. This can lead to “log corruption, output injection, and in some setups path traversal.”

4. Path Traversal File Write (CVE-2025-12972)

The most critical outcome is a path traversal vulnerability in the out_file plugin. When the file output lacks a defined File key, it uses the tag to create the file name. Since the tag can be controlled (and injected with ../ from the previous flaws), “Attackers can use path traversal characters ‘../’ in the Tag to change the file path and name.” This can lead to RCE on many systems, allowing attackers to “create or overwrite files anywhere on the filesystem, which can lead to tampered logs, planted malicious files, and remote code execution.”

5.  Stack Buffer Overflow (CVE-2025-12970)

Separately, a stack buffer overflow exists in the Docker Metrics input plugin (in_docker). The plugin copies a container’s name into a fixed 256-byte stack buffer without checking its length. An attacker who can create a container with an “excessively long name” can “crash the agent or execute arbitrary code.”

Fluent Bit users must take immediate action to secure their telemetry pipelines.

• Update Immediately: Update Fluent Bit to the latest stable version v4.1.1 or v4.0.12.
• Limit Dynamic Tags: “Prefer static, predefined tags in configurations to eliminate untrusted input influencing routing or file naming”.
• Lock Down Outputs: For the file output plugin, “explicitly set a fixed Path or File parameter to prevent tag-based path expansion or traversal”.
• Least-Privilege: Run Fluent Bit as a non-root user and restrict filesystem access.

Related Posts:

• Fluent Bit Vulnerabilities (CVE-2024-50608 & CVE-2024-50609) Could Disrupt Millions of Enterprise Systems
• Linguistic Lumberjack (CVE-2024-4323): Critical Vulnerability Shakes Cloud Logging Infrastructure
• Unauthenticated Attackers Can Hijack 400K+ WordPress Sites via Fluent Forms Bug (CVE-2024-2771)
• OpenAI Services Hit by Major Outage Due to Telemetry Service Deployment
• ByteDance’s Trae IDE Under Fire: AI Coding Tool Caught Telemetry Spying Even After Opt-Out