Summary
Qdrant versions 1.9.3 through 1.15.x contain an arbitrary file write issue that allows appending attacker-controlled data to files via the /logger endpoint.
An attacker with minimal (read-only) access can abuse the on_disk.log_file path to write to arbitrary locations on the server's filesystem, potentially impacting integrity and availability.
Affected context
This vulnerability affects Qdrant deployments that expose the /logger endpoint to users or services with read-only access.
The issue arises when the log configuration accepts an on_disk.log_file path from the client and uses it to open or append to files on disk without sufficient validation or restriction.
Any environment where this endpoint is reachable from untrusted or semi-trusted clients may be impacted, including internal microservice deployments if network boundaries are weak.
Exploit preconditions
- Attack surface: HTTP API, specifically the
/loggerendpoint. - Required access: Minimal authenticated or authorized access (read-only role) sufficient to call the logging configuration endpoint.
- Trigger point: Submitting a request that sets
on_disk.log_fileto an attacker-controlled filesystem path, causing Qdrant to append log data to that file. - Why exploitation is possible: The application trusts client-supplied file paths for log configuration and uses them directly on the filesystem without proper path validation, sandboxing, or access control checks.
Why this happens
The core issue is improper handling of externally controlled file paths, mapped to CWE-73: External Control of File Name or Path.
The /logger endpoint allows clients to influence logging configuration, including the on_disk.log_file path, and the server uses this value to open or append to files on disk.
Because the path is not restricted to an allowed directory, validated against a safe list, or otherwise constrained, an attacker can point logging to arbitrary filesystem locations and cause unintended writes.
Potential impact
- Appending malicious or crafted data to configuration or policy files, which may influence application or system behavior.
- Corrupting log files or other application data, impacting observability and forensic analysis.
- Leveraging file append to aid in other attacks (e.g., poisoning inputs, influencing downstream parsers, or interfering with backups).
- Degrading availability if critical files grow unexpectedly or are corrupted by appended data.
General mitigation guidance
Effective mitigation focuses on constraining how file paths from external inputs are interpreted and limiting where the application can write logs:
- Restrict log file paths to a predefined directory or set of directories, and reject any path outside those boundaries.
- Normalize and validate paths (e.g., remove
..segments, disallow absolute paths) before using them with filesystem APIs. - Prefer configuration-driven logging targets over runtime, user-controlled file path changes.
- Run the Qdrant process with least-privilege filesystem permissions so it cannot write to sensitive system or application directories.
- Consider using OS-level or container-level controls (e.g., read-only mounts, dedicated log volumes) to constrain where the process can write.
How teams detect issues like this
Detection generally involves both code-level and operational monitoring approaches:
- Static analysis / code review: Look for endpoints or configuration handlers that accept file paths from clients and pass them to logging or file APIs without strong validation or directory constraints.
- Configuration and policy checks: Enforce policies that logging destinations are fixed or drawn only from trusted configuration, not arbitrary user input.
- Runtime monitoring: Monitor log directories and sensitive paths for unexpected files or sudden growth correlated with
/loggerendpoint requests. - Operational auditing: Review access logs for unusual or frequent requests to the
/loggerendpoint or patterns indicating abuse of log configuration.

