CVE-2026-91771

Path traversal in wandb ≤ 0.28 allows arbitrary file writes via crafted filenames.

• CVSS 8.7
• 22
• Input Validation and Sanitization
• Remote

Weights & Biases wandb before 0.29.0 fails to validate the file name from server responses in the File.download function, allowing path traversal attacks. Attackers controlling the backend can supply file names with directory traversal sequences to write files outside the intended download directory, potentially enabling code execution through modification of shell startup files or Python import paths.

CVE

CVE-2026-91771

CWE

22

CVSS base score

8.7

Published

2026-09-15

OWASP

A05 Security Misconfiguration

Orthogonal defect classification

Checking

Code defect classification

Missing Check

Category

Input Validation and Sanitization

Subcategory

Path Traversal

Accessibility scope

Remote

Impact

Arbitrary Code Execution

Affected component

wandb

Fixed by upgrading

Yes

Solution

Upgrade Weights & Biases to version 0.29.0 or later, which validates file names and prevents path‑traversal downloads.

import os
import requests

class File:
    def __init__(self, url: str, download_dir: str):
        self.url = url
        self.download_dir = download_dir

    def download(self):
        # VULNERABLE: writes filename from server without validation, allowing path traversal
        resp = requests.get(self.url, stream=True)
        resp.raise_for_status()
        # server supplies header "X-Wandb-Filename"
        filename = resp.headers.get("X-Wandb-Filename", "download.bin")
        target_path = os.path.join(self.download_dir, filename)
        with open(target_path, "wb") as f:
            for chunk in resp.iter_content(chunk_size=8192):
                if chunk:
                    f.write(chunk)
        return target_path

import os
import requests

class File:
    def __init__(self, url: str, download_dir: str):
        self.url = url
        self.download_dir = download_dir

    def download(self):
        # FIX: sanitize filename to prevent directory traversal
        resp = requests.get(self.url, stream=True)
        resp.raise_for_status()
        filename = resp.headers.get("X-Wandb-Filename", "download.bin")
        # Strip path components and ensure the result stays within download_dir
        safe_name = os.path.basename(filename)
        target_path = os.path.normpath(os.path.join(self.download_dir, safe_name))
        # Ensure the normalized path is still under download_dir
        if not os.path.commonpath([self.download_dir, target_path]) == os.path.abspath(self.download_dir):
            raise ValueError("Unsafe filename detected")
        with open(target_path, "wb") as f:
            for chunk in resp.iter_content(chunk_size=8192):
                if chunk:
                    f.write(chunk)
        return target_path

__VAITP_MODEL_REFUSED__

@misc{vaitp:cve202691771,
  title        = {{Path traversal in wandb ≤ 0.28 allows arbitrary file writes via crafted filenames.}},
  author       = {Bogaerts, Fr\'ed\'eric and Ivaki, Naghmeh and Fonseca, Jos\'e},
  year         = {2026},
  note         = {VAITP Python Vulnerability Dataset, entry CVE-2026-91771},
  howpublished = {\url{https://netpack.pt/vaitp/vulnerability/CVE-2026-91771/}}
}