These pages are published after PatchSiren validates generated defensive summaries against stored public CVE and source evidence.
CVE-2026-65315 is a high-severity vulnerability in the Ollama project's GGUF metadata parser. The vulnerability allows remote attackers to crash the server by supplying a crafted GGUF file with attacker-controlled length and count fields. This vulnerability is triggered through the blob upload and model create or pull API endpoints. Successful exploitation requires the attacker to upload a sub-1KB crafted [truncated]
CVE-2026-15685 is a Denial-of-Service vulnerability in Ollama, caused by improper validation of array indices in the downloadBlob function. This vulnerability allows remote attackers to create a denial-of-service condition on affected installations of Ollama without requiring authentication. The vulnerability exists due to a lack of proper validation of user-supplied data, which can result in a memory acc [truncated]
CVE-2026-7482 is a heap out-of-bounds read in Ollama’s GGUF model loader that affects versions before 0.17.1. According to the CVE description and NVD record, a malicious GGUF file can declare tensor offsets and sizes that exceed the file’s actual length; during quantization in fs/ggml/gguf.go and server/quantization.go (WriteTo()), Ollama can read past the allocated heap buffer. The result is information [truncated]
## Summary Ollama for Windows versions 0.12.10 through 0.17.5 contain a path traversal vulnerability in their automatic update mechanism. The application constructs local file paths using attacker‑controlled HTTP response headers without validation, allowing directory traversal sequences (../) to be resolved. This enables an attacker who can influence update responses to write arbitrary files—including ex [truncated]
Ollama for Windows versions 0.12.10 through 0.17.5 fail to verify the integrity or authenticity of downloaded update executables. The Windows implementation of the update verification routine unconditionally returns success, bypassing digital signature and trust validation before staging or executing update payloads. Combined with silent automatic updates on Windows, this enables attacker-supplied executa [truncated]
A server-side request forgery vulnerability was found in Ollama up to 0.18.1. The issue affects the Model Pull API in the file server/download.go. The attack can be launched remotely. This vulnerability has a CVSS score of 5.3 and a severity of MEDIUM. Users and administrators should review the vulnerability details to understand the potential impact on their systems.
CVE-2025-15514 is a high-severity null pointer dereference vulnerability in Ollama's multi-modal model image processing functionality. The vulnerability allows remote attackers to exploit the application by sending specially crafted base64 image data, causing a segmentation fault and crashing the runner process, resulting in a denial of service condition. This issue affects Ollama versions from 0.11.5-rc0 [truncated]