Editor’s note: The following is a sponsored blog post from ReversingLabs.
Software supply chain security was first thrust into the cybersecurity world’s consciousness by SunBurst, the 2020 software supply chain attack on SolarWinds. Fast forward to 2025, and supply chain attacks are accelerating — and rapidly evolving, a new report by ReversingLabs (RL) finds.
The software industry’s growing reliance on artificial intelligence (AI) to generate code more quickly, coupled with widespread open-source software risks and a lack of visibility in commercial off-the-shelf (COTS) software, leave end user organizations exposed to damaging attacks. At the same time, outdated vulnerability management and application security (AppSec) tools are struggling to fill key detection gaps.
Here are some of the key findings from The 2025 Software Supply Chain Security Report from RL — as well as important takeaways for security leaders and teams.
Software supply chain risks grow
In RL’s previous supply chain security reports, malicious packages on open-source repositories like npm, the Python Package Index (PyPI) and RubyGems took center stage. Since 2022, researchers have seen a steady, upward trend of attackers leveraging these platforms to spread malware. The goal: Get unsuspecting developers to download and use their malicious code in their software.
However, while these kinds of attacks continued to plague open-source software (OSS) in 2024, RL found that simple, typosquatted malicious packages on these platforms declined by more than 70% from 2023 to 2024. In addition, discovered malware instances dropped by more than 85% on PyPI, alone.
RL researchers speculate that much of the decline in OSS threats is due to tightened security policies on platforms like PyPI. Those include developer mandates on the use of two-factor authentication (2FA) as well as closer monitoring of packages for malicious content. However, RL researchers stress that – despite these improvements – attacks on the supply chain have not gone away. If anything, they are growing in prominence, driven by continued security lapses in both OSS and closed-source, commercial software packages.
For example, while RL observed a decline in malicious packages, instances of leaked developer secrets such as hard-coded credentials, API and encryption keys jumped by 12% in the last year. Then there’s the vast quantities of closed-source, commercial software applications that form the foundation of most organizations’ IT infrastructure, which are black boxes when it comes to software security and integrity. An RL analysis of more than three dozen common, commercial binaries licensed to enterprises found clear evidence of insecure design and insufficient application hardening, as well as exposed data and development secrets across the majority of the applications.
Moving beyond vulnerabilities: Life after NVD
Identifying and tracking vulnerabilities via Common Vulnerabilities and Exposures (CVEs) has been a pillar of AppSec for decades. However, in 2024, this began to crumble. In February, the U.S. National Institute of Standards and Technology (NIST) announced that it would cease enriching CVEs.
NIST’s failure to enrich CVEs hobbled AppSec teams, denying them critical and actionable information such as severity scores (CVSS), patching statuses, vulnerability descriptions, and lists of affected products. This major change coincides with a breakdown in the CVE reporting system driven by an increased volume of CVEs and insufficient staffing and funding for the National Vulnerability Database (NVD), the U.S. government’s clearinghouse for CVEs.
This drop-off in the data needed for proper vulnerability management underscores the need for security leaders to rethink how their organizations should manage the security risks lying in their software products. RL researchers recommend going beyond vulnerability management to instead focus on the full array of risks to software supply chains, such as secrets exposures, tampering of the build environment and file rot.
To get the visibility they need, software risk managers need to re-evaluate their AppSec tooling. Technologies such as complex binary analysis and reproducible builds are key, the Enduring Security Framework (ESF) working group has urged.
AI development opens new doors for attackers
The use of AI codling tools exploded in the past two years as OpenAI’s ChatGPT, Microsoft’s Copilot and other generative AI tools have vastly expanded their capabilities. That includes the adoption of LLMs and generative AI by software development teams. This has opened new doors for cybercriminals, who are exploiting software vulnerabilities in the development infrastructure that is used to create and train AI models.
A common example of this can be seen with Python PICKLE files, which are popular program files used for object serialization. RL experts and others deem Pickle files as inherently unsafe, because they are designed as interpreters: blindly executing opcodes without first verifying the validity or integrity of the entire Pickle file. Practically, this allows embedded Python code to be run when the model is loaded onto a user’s machine. However, this opens the door for cyber criminals to execute malicious commands, inject malware onto affected systems, or engage in inbound and outbound communications with malicious infrastructure. Malicious actors also have the ability to influence (aka poison) large language models (LLMs) and generative AI with bad information or other kinds of attacks.
This rise in risk calls for enterprises to evaluate whether or not their AI use – whether it be for software development or for the software products they are purchasing – is responsible. That requires vetting the security of ML models with the same rigor as any other software package.
Take the next step
Considering the growing and evolving nature of software supply chain threats, security professionals and risk management teams alike need to adjust their strategies for protecting their enterprises for 2025.
To learn more about what kinds of strategies will be most effective, download The 2025 Software Supply Chain Security Report for a deep dive into the software threat landscape. Also, be sure to watch RL’s webinar about the report.
