CVE-2025-0315:
Ollama Allocation of Resources Without Limits or Throttling vulnerability
7.5
CVSS Score
3.0
Basic Information
CVE ID
GHSA ID
EPSS Score
0.19664%
CWE
Published
3/20/2025
Updated
3/21/2025
KEV Status
No
Technology
Go
Technical Details
CVSS Vector
CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Package Name | Ecosystem | Vulnerable Versions | First Patched Version |
---|---|---|---|
github.com/ollama/ollama | go |
Vulnerability Intelligence
Miggo AI
Root Cause Analysis
The vulnerability stems from processing untrusted GGUF model files during creation/loading. The CreateModel endpoint (ModelHandler) would be the entry point for accepting malicious uploads, while LoadGGUF (GGUF parser) would handle the actual memory allocation based on file metadata. Both lack resource limits: CreateModel doesn't enforce upload quotas, and LoadGGUF trusts the file's tensor size declarations without validation. This matches CWE-770's pattern of missing throttling/limits during resource allocation.