VAITP Dataset

← Back to the dataset

CVE-2026-119120

Crawl4AI lets auth clients pick malicious HuggingFace models via unfiltered API, leading to code exec or resource exhaustion.

  • CVSS 7.7
  • CWE-502
  • Input Validation and Sanitization
  • Remote

Crawl4AI from 0.9.0 before 0.9.5 fails to filter the CosineStrategy model_name field in untrusted Docker API request bodies, allowing authenticated clients to choose which HuggingFace model the server loads. Attackers can reference a malicious repository with a pickle weight file to execute Python code under PyTorch below 2.6, or exhaust disk and memory.

CVSS base score
7.7
Published
2026-10-29
OWASP
A08 Software and Data Integrity Failures
Orthogonal defect classification
Interface
Code defect classification
Incorrect Interface
Category
Input Validation and Sanitization
Subcategory
Insecure Parsing or Deserialization
Accessibility scope
Remote
Impact
Arbitrary Code Execution
Affected component
Crawl4AI
Fixed by upgrading
Yes

Solution

Upgrade Crawl4AI to version 0.9.5 or later, which adds proper validation of the CosineStrategy model_name field.

Vulnerable code sample

import json
import torch
from transformers import AutoModel

def load_model_from_request(request_body: bytes) -> torch.nn.Module:
    """
    Parse a Docker API request body and load the requested HuggingFace model.
    """
    data = json.loads(request_body)
    # VULNERABLE: unsanitized model_name allows loading arbitrary repository
    model_name = data.get("model_name", "bert-base-uncased")
    # Directly pass user‑controlled name to AutoModel.from_pretrained
    model = AutoModel.from_pretrained(model_name)
    return model

Patched code sample

import json
import torch
from transformers import AutoModel

ALLOWED_MODELS = {"bert-base-uncased", "distilbert-base-uncased"}

def load_model_from_request(request_body: bytes) -> torch.nn.Module:
    """
    Parse a Docker API request body and load the requested HuggingFace model.
    """
    data = json.loads(request_body)
    # FIX: validate model_name against whitelist before loading
    requested = data.get("model_name", "bert-base-uncased")
    model_name = requested if requested in ALLOWED_MODELS else "bert-base-uncased"
    model = AutoModel.from_pretrained(model_name)
    return model

Cite this entry

@misc{vaitp:cve2026119120,
  title        = {{Crawl4AI lets auth clients pick malicious HuggingFace models via unfiltered API, leading to code exec or resource exhaustion.}},
  author       = {Bogaerts, Fr\'ed\'eric and Ivaki, Naghmeh and Fonseca, Jos\'e},
  year         = {2026},
  note         = {VAITP Python Vulnerability Dataset, entry CVE-2026-119120},
  howpublished = {\url{https://netpack.pt/vaitp/vulnerability/CVE-2026-119120/}}
}

Download this entry as JSON Download the whole dataset

Introducing the "VAITP dataset": a specialized repository of Python vulnerabilities and patches, meticulously compiled for the use of the security research community. As Python's prominence grows, understanding and addressing potential security vulnerabilities become crucial. Crafted by and for the cybersecurity community, this dataset offers a valuable resource for researchers, analysts, and developers to analyze and mitigate the security risks associated with Python. Through the comprehensive exploration of vulnerabilities and corresponding patches, the VAITP dataset fosters a safer and more resilient Python ecosystem, encouraging collaborative advancements in programming security.

The supreme art of war is to subdue the enemy without fighting.

Sun Tzu – “The Art of War”

:: Shaping the future through research and ingenuity ::

Before you use the VAITP AI models

Please read and accept the following disclaimer:

By choosing "I understand and accept", you acknowledge that you understand and agree to these terms.