Skip to content
supplychainattack.orgSupply chain attack incident catalog
resolvedcritical

Malicious code in deepface-weight (PyPI)

The PyPI package deepface-weight contained malicious code that exfiltrated Telegram session files on import, granting attackers full access to associated Telegram accounts. The package was a typosquat of the legitimate deepface ML library with no actual machine-learning functionality.

ShareXLinkedInHacker News
Disclosed
Last updated
Blast radius
Any user who installed and imported the deepface-weight package; Telegram accounts associated with telethon session files in the installer's working directory.
Ecosystems
Attack vectors
Affected entities
  • deepface-weightMalicious PyPI package mimicking the legitimate deepface library

The deepface-weight package published to PyPI contained intentionally malicious code designed to steal Telegram authentication credentials. Upon import, the package spawned a background daemon thread that polled the installer's working directory for data/telethon_market_userbot.session files for approximately 10 minutes and exfiltrated them via HTTP POST to a hardcoded webhook.site endpoint (https://webhook.site/d6ea9c5b-4a85-4e59-9397-2bb5f9407c87).\n\nTelethon session files contain live authentication material that grants full access to the associated Telegram account. The malicious code included a variable explicitly named evil_server_url with a Russian-language comment identifying it as the attacker's server, indicating deliberate intent.\n\nThe package was a typosquat of the popular deepface machine-learning library but contained no ML code. Author metadata consisted of placeholder text (asdqwdasdqwdasd) and a disposable email address at playboot.com, further indicating malicious intent.\n\nThe incident was identified and reported through the OpenSSF malicious-packages repository (MAL-2026-14158) and subsequently published as GitHub Advisory GHSA-5j48-33h2-gvwg.

Indicators of compromise

Packages
  • deepface-weight
Domains
  • webhook.site

Remediation

  • Immediately uninstall deepface-weight from all systems where it was installed
  • Revoke or reset any Telegram accounts that may have had session files in the installer's working directory
  • Audit logs for any unauthorized Telegram account activity during the period the package was installed
  • Use the legitimate deepface package (from PyPI) instead of deepface-weight
  • Review PyPI package names carefully before installation to avoid typosquats
  • Monitor for similar typosquatting attempts targeting popular ML libraries

Sources

  1. GitHub Advisory GHSA-5j48-33h2-gvwg · GitHub Advisory Database

Cite this entry

"Malicious code in deepface-weight (PyPI)." supplychainattack.org, Supply Chain Attack Incident Catalog. Disclosed August 19, 2026; last updated August 19, 2026. https://supplychainattack.org/incident/malicious-code-in-deepface-weight-pypi-k5ay6t

Suggest a correction

Found an error or have a newer source? Corrections to factual errors take priority over new entries.

  1. containedcritical

    Malicious code in scrambleeeer (PyPI)

    The PyPI package scrambleeeer contains malicious code that establishes a reverse shell to a hardcoded location, allowing remote command execution on affected systems. The package was identified as part of a malicious campaign and has been documented by the OpenSSF.

    2026 08 ScrambleeerPyPICompromised package
  2. containedcritical

    Malicious code in reqcrypts (PyPI)

    The reqcrypts package on PyPI contains malicious code that implements a hidden backdoor. The package masquerades as an HTTP request library but secretly monitors responses for specific fields and executes their content without user knowledge.

    2026 08 ReqcryptPyPICompromised packageMalicious commit
  3. containedcritical

    Malicious code in boto4 (PyPI)

    A malicious package named boto4 was published to PyPI containing embedded executable code capable of cryptomining, remote command execution, persistence, data exfiltration, and worm-style propagation controlled via Telegram bot. The package was identified and attributed to the 2026-08-boto4 campaign by the OpenSSF.

    2026 08 Boto4PyPICompromised package
  4. containedcritical

    Malicious code in scrambleeer (PyPI)

    The scrambleeer package on PyPI contained malicious code that establishes a reverse shell to a hardcoded location, enabling arbitrary command execution on affected systems. The malicious package was identified and cataloged by the OpenSSF malicious-packages project.

    2026 08 ScrambleeerPyPICompromised package