CVE-2026-112107
Received Received - Intake

Buffer Overflow in FastLLM Safetensors Weight Loading

Publication date: 2026-10-19

Last updated on: 2026-10-19

Assigner: VulDB

Description
A flaw has been found in ztxz16 fastllm up to 1.8.2. Affected by this issue is the function SafeTensorItem::CreateBuffer of the file src/model.cpp of the component Safetensors Weight Loading. This manipulation of the argument data_offsets causes buffer overflow. The attack can be initiated remotely. The exploit has been published and may be used. The project was informed of the problem early through an issue report but has not responded yet.

CVSS Scores

EPSS Scores

Probability:
Percentile:

Meta Information

Published
2026-10-19
Last Modified
2026-10-19
Generated
2026-10-19
AI Q&A
2026-10-19
EPSS Evaluated
N/A
NVD
CVE-2026-112107
EUVD
EUVD-2026-104243

Affected Vendors & Products

Showing 3 associated CPEs
Vendor Product Version / Range
ztxz16 fastllm 1.8.0
ztxz16 fastllm 1.8.1
ztxz16 fastllm 1.8.2

Helpful Resources

Exploitability

CWE
CWE Icon
KEV
KEV Icon
CWE ID Description
CWE-120 The product copies an input buffer to an output buffer without verifying that the size of the input buffer is less than the size of the output buffer.
CWE-119 The product performs operations on a memory buffer, but it reads from or writes to a memory location outside the buffer's intended boundary. This may result in read or write operations on unexpected memory locations that could be linked to other variables, data structures, or internal program data.

Attack-Flow Graph

AI Quick Actions

Instant insights powered by AI
Executive Summary

This vulnerability is a heap buffer overflow in the fastllm library's safetensors weight loading mechanism. It occurs when the declared shape of a tensor and its data offsets in the safetensors header are inconsistent, leading to an out-of-bounds write. The SafeTensorItem constructor computes the expected buffer length from the tensor's shape but allocates memory based on the data offsets without validating their consistency.

Detection Guidance

To detect this vulnerability, inspect systems running fastllm versions up to 1.8.2 for suspicious model loading activity. Check for crashes or memory corruption errors during model inference. Use tools like AddressSanitizer or Valgrind to monitor for heap buffer overflows when loading Safetensors files. Examine logs for malformed Safetensors headers or inconsistent tensor shapes and data offsets.

Impact Analysis

This vulnerability can allow attackers to overwrite adjacent heap objects with malicious data, potentially leading to arbitrary code execution. It affects any deployment loading untrusted Hugging Face-style model directories via fastllm's CLI, HTTP server, or Python binding. The exploit can be initiated remotely.

Compliance Impact

This vulnerability could potentially impact compliance with GDPR and HIPAA by enabling remote code execution through buffer overflow attacks. If exploited, it may lead to unauthorized access, data breaches, or data integrity issues, violating confidentiality and integrity requirements under these regulations.

Mitigation Strategies

Immediately upgrade fastllm to a patched version if available. If not, avoid loading untrusted Safetensors models. Implement input validation to ensure tensor shapes match data offsets before processing. Disable network-exposed model loading interfaces until fixed. Monitor for exploitation attempts using network traffic analysis for malformed Safetensors payloads.

Chat Assistant

Ask questions about this CVE
Hi! I’m here to help you understand CVE-2026-112107. Ask me anything about the vulnerability, its impact, or mitigation strategies.
0/70

EPSS Chart