Langflow Unauthenticated File Upload Leads to DoS and Information Leak (CVE-2026-55450)

A critical vulnerability, CVE-2026-55450, has been identified in Langflow, affecting all versions prior to 1.9.1. This flaw allows unauthenticated attackers with network access to a Langflow instance to upload files of any size to the server's local filesystem through the deprecated /api/v1/upload/{flow_id} endpoint. The vulnerability stems from a lack of authentication and file size validation on this endpoint. Exploitation can lead to severe resource exhaustion, specifically disk space depletion, which results in a Denial-of-Service (DoS) condition for the Langflow application and potentially the host system. Furthermore, the server's response to successful uploads includes the absolute path of the saved file, disclosing sensitive system information that could aid in chaining further attacks. This vulnerability was reported via GitHub Advisory GHSA-x223-p2gf-v735.

Attack Chain

1. An unauthenticated attacker identifies a publicly accessible Langflow instance running a vulnerable version (prior to 1.9.1).
2. The attacker sends an HTTP POST request to the deprecated and unauthenticated /api/v1/upload/{flow_id} endpoint (e.g., http://langflow_host:7860/api/v1/upload/A_VALID_UUID) using a tool like curl.
3. The request includes a large file as multipart/form-data, specified via the -F "file=@<any_file>" parameter.
4. Due to the absence of authentication checks and file size validation on this endpoint, the Langflow application's save_uploaded_file function processes and writes the attacker's large file directly to the local filesystem, within a cache directory structure derived from the provided flow_id.
5. The attacker repeatedly performs these large file uploads, consuming increasing amounts of disk space on the Langflow server.
6. Upon each successful upload, the server returns an HTTP 201 Created status, and the response body contains the absolute path to the newly saved file on the server (e.g., /Users/ori/Library/Caches/langflow/A_VALID_UUID/filename.txt).
7. The continuous and uncontrolled writing of large files leads to the exhaustion of available disk space, resulting in a Denial-of-Service condition for the Langflow application and potentially critical system functions.
8. The leaked absolute file paths expose details about the server's directory structure, which can be leveraged by the attacker for subsequent reconnaissance or to craft more sophisticated attacks like path traversal or local file inclusion.

Impact

The primary impact of CVE-2026-55450 is the potential for a complete Denial-of-Service (DoS) for the affected Langflow application and its host system. Unauthenticated attackers can exhaust server disk space by continuously uploading large files, rendering the application inoperable and potentially destabilizing the entire server. Additionally, the vulnerability leads to an information leak, where absolute file paths on the server are disclosed in response to file uploads. While the source does not provide victim numbers or specific targeted sectors, any organization deploying Langflow versions prior to 1.9.1 is at risk of operational disruption and reconnaissance from this flaw. The leaked path information could facilitate further, more targeted attacks.

Recommendation

• Patch CVE-2026-55450 immediately: Upgrade Langflow installations to version 1.9.1 or later to mitigate the unauthenticated upload and disk exhaustion vulnerabilities.
• Deploy the Sigma rules in this brief: Implement Detects CVE-2026-55450 Exploitation - Unauthenticated Langflow Upload and Detects CVE-2026-55450 Exploitation - Successful Unauthenticated Langflow Upload into your SIEM to detect attempted and successful exploitation.
• Enable comprehensive web server logging: Ensure detailed HTTP access logs are enabled for the web server hosting Langflow, capturing cs-method, cs-uri-stem, sc-status, and c-ip to effectively utilize the detection rules provided.
• Monitor disk space utilization: Implement alerts for critical disk space thresholds on systems hosting Langflow to identify potential DoS attacks.