CVE-2024-10821

7.5 HIGH

📋 TL;DR

This CVE describes a Denial of Service vulnerability in Invoke-AI server v5.0.1 where unauthenticated attackers can send specially crafted multipart requests to the /api/v1/images/upload endpoint, causing the server to enter an infinite loop and consume excessive resources. This affects all users of vulnerable Invoke-AI servers, potentially rendering the service completely unavailable.

💻 Affected Systems

Products:
  • Invoke-AI
Versions: v5.0.1
Operating Systems: All platforms running Invoke-AI server
Default Config Vulnerable: ⚠️ Yes
Notes: Only affects the Invoke-AI server when the image upload endpoint is enabled and accessible.

⚠️ Manual Verification Required

This CVE does not have specific version information in our database, so automatic vulnerability detection cannot determine if your system is affected.

Why? The CVE database entry doesn't specify which versions are vulnerable (no version ranges provided by the vendor/NVD).

🔒 Custom verification scripts are available for registered users. Sign up free to download automated test scripts.

Recommended Actions:
  1. Review the CVE details at NVD
  2. Check vendor security advisories for your specific version
  3. Test if the vulnerability is exploitable in your environment
  4. Consider updating to the latest version as a precaution

⚠️ Risk & Real-World Impact

🔴

Worst Case

Complete service outage for all users, requiring server restart and potentially causing data loss or corruption.

🟠

Likely Case

Temporary service disruption affecting all users until the server is restarted or the attack stops.

🟢

If Mitigated

Minimal impact with proper rate limiting, request validation, and monitoring in place.

🌐 Internet-Facing: HIGH - The vulnerable endpoint is accessible without authentication and can be exploited remotely.
🏢 Internal Only: MEDIUM - Internal attackers could still cause service disruption, but network segmentation reduces exposure.

🎯 Exploit Status

Public PoC: ✅ No
Weaponized: UNKNOWN
Unauthenticated Exploit: ⚠️ Yes
Complexity: LOW

Exploitation requires sending specially crafted multipart requests with excessive characters in boundary parameters.

🛠️ Fix & Mitigation

✅ Official Fix

Patch Version: v5.0.2 or later

Vendor Advisory: https://huntr.com/bounties/0ac24835-c4c0-4f11-938a-d5641dfb80b2

Restart Required: Yes

Instructions:

1. Backup current configuration and data. 2. Stop the Invoke-AI server. 3. Update to v5.0.2 or later using your package manager or from GitHub releases. 4. Restart the server. 5. Verify the fix by testing the upload endpoint.

🔧 Temporary Workarounds

Rate Limiting

all

Implement rate limiting on the /api/v1/images/upload endpoint to prevent excessive requests.

# Configure rate limiting in your web server or application firewall

Request Validation

all

Add middleware to validate multipart request boundaries and reject malformed requests.

# Implement boundary validation in your application code

🧯 If You Can't Patch

  • Implement network-level controls to restrict access to the vulnerable endpoint
  • Deploy a WAF with rules to detect and block malformed multipart requests

🔍 How to Verify

Check if Vulnerable:

Check if running Invoke-AI server version v5.0.1 and the /api/v1/images/upload endpoint is accessible.

Check Version:

invokeai --version or check package.json/version file in installation directory

Verify Fix Applied:

After patching, test the upload endpoint with normal and malformed requests to ensure it handles boundaries correctly without entering infinite loops.

📡 Detection & Monitoring

Log Indicators:

  • Excessive CPU/memory usage spikes
  • Multiple failed upload attempts with malformed boundaries
  • Server process hanging or restarting frequently

Network Indicators:

  • Unusually large multipart requests to /api/v1/images/upload
  • Multiple rapid requests from single IPs

SIEM Query:

source="invokeai.log" AND ("upload" OR "/api/v1/images/upload") AND ("timeout" OR "hang" OR "high_cpu")

🔗 References

📤 Share & Export