Site navigation

Open Source Malware Spikes 140% in Q3’25

Tom Quinn

,

open source malware
Attackers are increasingly using multi-stage malware to embed backdoors and info-stealers in widely used dependencies, threatening the open source software supply chain.

Incidents of open source malware have surged by 140% in the third quarter of 2025, according to new figures from Sonatype, a record rise fuelled by sophisticated, stealth-first attacks targeting open source registries like npm.

Sonatype’s latest Open Source Malware Index details the more than 34,300 new open source malware packages identified by the software platform, with this quarter seeing a trend in supply chain attacks exposing new weak links in open source security.

A series of npm attacks illustrates this dangerous escalation. Sonatype recently observed the compromise of widely used npm packages chalk, debug, and more than a dozen other such dependencies, which see more than 2 billion downloads every week.

These packages, which other software relies on to function, are often included automatically when developers install larger frameworks or tools, meaning that attackers can effectively poison the software supply chain by injecting malicious code into tools that thousands (or millions) of developers use, and distribute malware at scale.

Meanwhile, Sonatype also witnessed the unprecedented Shai-Hulud campaign, which exhibited ‘worm-like’ behaviour that allowed malicious code to self-propagate across repositories, exfiltrate credentials, and publish new compromised packages.

This attack uses a multi-step payload designed to steal credentials from a developer’s machine. If it then gains access to GitHub, it then publishes the repository Shai-Hulud, uploads the stolen secrets there, and spreads further by injecting malicious code into other npm packages used in the same project.

Sonatype said that, so far, collective efforts have tracked more than 180 npm packages compromised with Shai-Hulud.

Another disturbing trend noted in the report is attackers’ increasing use of multi-stage and stealth-first attacks. Droppers, for instance, which act as lightweight delivery mechanisms that install secondary payloads such as backdoors or info-stealers, skyrocketed in Q3, making up nearly 38% of all threats, while backdoor-laden packages grew 143% quarter-over-quarter.

Sonatype said this indicates a strategic evolution in adversaries building multi-stage malware that installs, hides, and maintains long-term access, posing as benign dependencies.

But what do these attackers hope to gain? According to Sonatype, data is the ultimate target, with data exfiltration malware accounting for 37% of all malicious open source packages detected over the quarter.

The report notes a growing trend toward threat actors engaging in intelligence-gathering, espionage, and monetisation of stolen data, with adversaries targeting developer credentials, access tokens, and proprietary information.


Recommended reading


This is transforming open source ecosystems into rich hunting grounds for data-driven exploitation, with stealthy attackers embedding malware into widely used code in the hope of striking lucky.

“The era of noisy, opportunistic malware is over. Attackers are patient, organised, and increasingly using AI to embed themselves inside the very tools developers rely on,” said Brian Fox, CTO and co-founder of Sonatype. 

“They’re hiding malicious payloads in plain sight, turning trusted open source dependencies into delivery mechanisms for data theft and persistence. Defenders need to match that sophistication with AI-driven visibility and proactive controls that stop threats before they ever reach a developer’s environment.”

Tom Quinn

Staff Writer, DIGIT

Latest News

AI

Nvidia Launches Open Secure AI Alliance for AI Safety and Security

AI Business Recruitment

Nearly a Quarter of Orgs Reducing Entry-level Hiring Due to AI Automation

Business

Scottish Businesses Turn to Self-funding as Growth Confidence Dips in H2

Data Finance

Payment Leaders are Struggling to Get Real-time Data