CVE-2024-12063 Overview
CVE-2024-12063 is a Denial of Service (DoS) vulnerability affecting imartinez/privategpt version 0.6.2. The flaw resides in the file upload feature, which improperly handles form-data containing oversized filenames. An attacker can submit a crafted multipart request with an excessively large filename value, exhausting server resources and rendering the application unavailable to legitimate users. The vulnerability is remotely exploitable over the network without authentication or user interaction. It is categorized under CWE-400 (Uncontrolled Resource Consumption) and affects the confidentiality-preserving local LLM deployment tool used in enterprise AI workflows.
Critical Impact
Unauthenticated remote attackers can crash PrivateGPT instances by sending a single malformed file upload request with an oversized filename, disrupting AI service availability.
Affected Products
- Pribai PrivateGPT version 0.6.2
- imartinez/privategpt file upload component
- Deployments exposing the PrivateGPT web interface to untrusted networks
Discovery Timeline
- 2025-03-20 - CVE-2024-12063 published to the National Vulnerability Database (NVD)
- 2026-06-17 - Last updated in NVD database
Technical Details for CVE-2024-12063
Vulnerability Analysis
The vulnerability stems from unbounded resource consumption during multipart form-data parsing in the PrivateGPT file upload endpoint. When a client submits a file upload request, the server processes the filename field within the Content-Disposition header without enforcing a maximum length constraint. Supplying an excessively large filename forces the application to allocate memory and CPU cycles disproportionate to the request size. This behavior aligns with the [CWE-400] classification for uncontrolled resource consumption.
Because PrivateGPT is often deployed to process sensitive documents in local AI pipelines, a successful attack disrupts document ingestion and retrieval-augmented generation (RAG) workflows. Repeated requests can sustain the outage indefinitely, denying access to all users of the affected instance.
Root Cause
The root cause is missing input validation on the filename parameter within the multipart form-data handler. The upload logic does not enforce length limits, reject malformed metadata, or apply resource quotas before parsing the request body. This allows a single request to consume server memory until the process becomes unresponsive.
Attack Vector
An unauthenticated attacker reaches the file upload endpoint over the network and issues an HTTP POST request. The request contains a multipart form-data payload with a filename field of extreme length. The server attempts to process the header value, exhausting available memory or CPU. Additional details are available in the Huntr Bug Bounty Listing.
Detection Methods for CVE-2024-12063
Indicators of Compromise
- Inbound HTTP POST requests to PrivateGPT upload endpoints with Content-Disposition filename values exceeding several kilobytes
- Sudden spikes in memory or CPU utilization on hosts running privategpt version 0.6.2
- Unresponsive PrivateGPT service processes or repeated worker restarts in application logs
- Multiple malformed multipart requests originating from a single source IP address
Detection Strategies
- Deploy web application firewall (WAF) rules that inspect multipart form-data headers and reject requests where the filename field exceeds a reasonable threshold, such as 255 characters
- Instrument the PrivateGPT process with resource monitoring to alert on abnormal memory growth tied to inbound upload requests
- Correlate HTTP access logs with process crashes or out-of-memory events to identify exploitation attempts
Monitoring Recommendations
- Track request size distributions and header lengths at the reverse proxy layer to establish a baseline for anomaly detection
- Alert on repeated HTTP 5xx responses from the /upload route or its equivalent within the PrivateGPT API
- Monitor system-level metrics for sustained memory pressure or process termination events on AI application servers
How to Mitigate CVE-2024-12063
Immediate Actions Required
- Restrict network access to PrivateGPT instances so that only trusted clients can reach the upload endpoint
- Place PrivateGPT behind a reverse proxy that enforces strict header and body size limits before requests reach the application
- Audit deployments to identify all hosts running privategpt version 0.6.2 and prioritize them for remediation
Patch Information
No vendor advisory or official patch is referenced in the enriched CVE data at the time of publication. Consult the upstream imartinez/privategpt repository for the latest releases and review the Huntr Bug Bounty Listing for remediation status. Until a fixed version is confirmed, apply the workarounds below.
Workarounds
- Configure the reverse proxy (for example, nginx client_max_body_size and large_client_header_buffers) to reject oversized multipart headers before they reach PrivateGPT
- Require authentication in front of the PrivateGPT interface using an identity-aware proxy or VPN gateway
- Implement per-source rate limiting on upload endpoints to blunt repeated exploitation attempts
- Run PrivateGPT with strict memory cgroup limits so that a single malformed request cannot exhaust host resources
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
