Malware surge in open source software alarms firms

Malware in open-source software rose sharply in 2025, according to Endor Labs, which found that more than 90% of open-source malware advisories over the past two years were reported last year.

The study combined a survey of more than 600 IT professionals with an analysis of Open Source Vulnerability database records and npm package data. It found that 51% of packages analysed in 2025 contained malware, while the number of advisories was 13.6 times higher than in January 2024.

The findings highlight growing concern among companies that rely on open source components in software development. While 81% of security leaders identified malicious open source software as a top priority for 2026, only 48% planned to increase budgets to address it.

The research also found a gap between awareness and operational controls. Although 88% of respondents said the first few days after a package release were the highest-risk period, only 21% said their organisations enforced protections such as cooldown periods before newly released packages could be used.

Supply Chain Risk

The report also tracked compromised maintainer accounts on npm, one of the largest software package repositories. It found that 92% of npm account takeovers in its dataset occurred in 2025, suggesting that attackers are increasingly targeting trusted developer accounts rather than publishing entirely new malicious packages.

This tactic can make malicious code harder to detect. Malware can be triggered as soon as a package is installed, allowing it to evade some traditional detection methods that focus on code execution at a later stage or on known vulnerability patterns.

Endor Labs argued that many security programmes still treat open-source malware as a series of isolated incidents rather than a broader supply chain problem. That, it said, leaves organisations exposed when malicious versions are automatically downloaded into development environments before teams have time to review them.

Its analysis also found that many compromised packages remained available for download after being reported. Only 14% of previously compromised npm packages in the sample used Trusted Publishing, a control intended to reduce the risk of account abuse and unauthorised package releases.

Budget Pressure

The figures suggest companies are still weighing the threat against broader spending demands. While awareness appears high, fewer than half of respondents said they expected their organisation to commit more money to addressing malicious open-source software in the coming year.

Responsibility for managing the risk is also spread across multiple functions. Survey respondents came from DevOps, security, and software engineering roles, reflecting that oversight of open-source software often spans teams rather than a single owner.

That fragmentation can slow decision-making when malicious packages are discovered. In fast-moving package ecosystems, even short-lived malicious versions can be downloaded into large numbers of environments within hours, according to the report.

The research was based on organisations with more than 100 employees and was conducted at a 95% confidence level with a margin of error of plus or minus 4%. The results reflected a global sample of IT professionals.

Another concern highlighted in the study was the effect of newer software tools on the attack surface. The report linked the rise in malicious packages to changes in software development practices, including the use of AI-related tools and dependencies.

Varun Badhwar, chief executive officer of Endor Labs, said current security approaches have not kept pace with the way attackers now operate in open-source ecosystems. "Most application security programs were built around vulnerability management, not to detect malware in the software supply chain. Attackers understand this. AI coding agents, MCP servers, and model dependencies are creating new entry points, and we're already seeing an uptick in malware in open source ecosystems targeting AI coding agents."

He added that the speed of attacks was outstripping the pace of organisational response. "The gap between how fast attackers move and how fast organisations respond is widening, and without a coordinated, cross-functional approach, even strong controls fail in practice."