Skip to content
Malware Campaign Exploits Hugging Face for Crypto Mining

Malware Campaign Exploits Hugging Face for Crypto Mining

First seen 29 Oct 2026, 22:44 UTC • •
API

Article Content

Browse articles
ThreatCluster AI
ThreatCluster •October 30, 2026 at 00:14 UTC
  • •The campaign exploits Hugging Face's 'trust_remote_code' feature to run malicious code.
  • •Over 32,000 downloads of malicious models were recorded, with significant GPU hijacking impact.
  • •New repositories emerged within days of takedown, indicating ongoing threat actor activity.

A cryptomining and backdoor campaign has resurfaced, exploiting Hugging Face model repositories to hijack GPUs. Researchers at Pluto Security identified 31 malicious repositories across seven accounts, with the most popular model downloaded over 28,000 times. The attack leverages the 'trust_remote_code' option, allowing malicious Python code to execute on user machines upon model loading. This campaign has recorded more than 32,000 downloads collectively, targeting GPU hosts running self-hosted inference servers like Xinference and GPUStack. Despite previous takedowns, new repositories appeared within days, indicating the persistence of the threat. Scanning tools failed to detect the malicious code, as it does not match known malware signatures. The campaign's operator has reportedly earned around $5,000 in cryptocurrency from the mining activities.

Start a free Starter trial for enhanced analysis

Ask AI about this cluster

Updated just now How this analysis works

Timeline

2024-05-10
CVE-2024-34359 published
Vulnerability assigned a CVE identifier and published in the National Vulnerability Database.
MITRE
2026-01-21
CVE-2026-22807 published
Vulnerability assigned a CVE identifier and published in the National Vulnerability Database.
MITRE
2026-10-29
Pluto Security reports on Hugging Face malware campaign
Pluto Security details a cryptomining campaign exploiting Hugging Face repositories, highlighting the use of trust_remote_code.
Thenextweb
2026-10-29
BleepingComputer reports on malware campaign return
BleepingComputer confirms the resurgence of the Hugging Face malware campaign, noting over 28,000 downloads of the most popular malicious model.
Bleepingcomputer
2026-10-29
Pluto Security publishes detailed analysis
Pluto Security's blog provides an in-depth analysis of the malware campaign, including indicators for defenders.
pluto.security

More articles in this cluster (12)

Following this threat?

Track NetSupport RAT and CVE-2024-34359 in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.

Free account · no card needed

Extracted Entities (22)

1 / 4

Common questions

How does the malware execute?
The malware executes by using the 'trust_remote_code' feature, allowing it to run Python code on the user's machine as soon as the model loads.
What systems are primarily targeted?
The campaign targets GPU hosts running self-hosted inference servers, particularly Xinference and GPUStack.
What should organizations do to protect themselves?
Organizations should avoid enabling the 'trust_remote_code' option and monitor for unusual activity related to GPU usage.