What is the JFrog Software Supply Chain Security State of the Union report?

Now in its fourth year, this annual research report tracks software supply chain threats, open source vulnerabilities, CVE trends, and more. It's designed to show where software supply chain security stands today, where governance is failing, and where attackers are likely to focus next.

How is this different from other security reports?

The JFrog Software Supply Chain Security Report for 2026 employs a three-pillar methodology, utilizing expert intelligence on a massive scale. It includes original threat research from the JFrog Security Research team, direct and anonymized data from thousands of real-world enterprises, and a commissioned survey of 1,500+ professionals across eight countries.

Who is this report for?

Security engineers, DevOps and platform teams, and industry leaders who need to know where real risk is accumulating, not just what the CVE databases say. It's also built for anyone who needs to justify investment to a board: the data shows not just what to fix, but why it matters and what peers are doing about it.

Does the report address AI security risks in the software supply chain?

Yes. As organizations adopt AI models, they introduce new attack vectors. We examine the specific dangers of running AI on internal infrastructure, including emerging security risks. Read the report to understand how to govern these new dependencies effectively and manage the AI security risks that software supply chain teams face today.

How are malicious packages in 2026 impacting development?

Attackers are finding new ways to infiltrate registries, leading to a rise in sophisticated campaigns. Our research focuses on the surge in these threats and the ways they bypass traditional security layers—using npm as a primary example of these evolving tactics. The full report details the volume and impact of these attacks on modern development cycles.

JFrog 2026年软件供应链安全现状报告

日益加剧的治理鸿沟。

更庞大的代码包体量、更多潜藏恶意的 AI 模型，以及一年前甚至根本不曾存在的全新攻击面。然而，在这重重风暴的中心，各大组织却将他们所使用的应用安全工具数量砍掉了近乎一半。企业深信自己正在变得更精简、更高效、更具目的性、掌控力更强，并且成功实现了“用更少的资源做更多的事”。但事实真的如此吗？

2026年度报告汇集了 JFrog 安全研究团队 JFrog Security Research, 的原创威胁情报、来自 JFrog 全球平台的整年运行数据，以及遍布全球 8 个国家的 1,500 多名安全和 DevOps 专业人员的真实反馈。本报告旨在揭示当今软件供应链安全的真实境况，并直面那些正在以超出所有人想象的速度疯狂蔓延的治理鸿沟。

下载报告

看不清验证码？点击刷新

《JFrog 2026年软件供应链安全报告》核心发现：

2025年，进入软件供应链的新增开源包数量同比激增 67%。

审查了多少？

2025年有大量全新的恶意包被揭露，然而在这一年里，全球仅有 40% 的组织部署了相应的检测工具。

哪个包位居榜首？

的组织选择在自己的基础架构上运行 AI 模型，这让他们面临着开源模型内部潜藏恶意载荷（Payloads）的安全风险。

里面是什么？

的组织在面对审计人员的问询时，需要花费超过一周的时间，才能证明单个应用程序是符合安全与合规要求的。

差距在哪里？

生态版图正在被重新改写

在企业级包生态系统中，npm 的流量已正式超越 Maven，跃升为使用最广泛的生态系统。与此同时，Hugging Face 上的新模型数量已突破 140 万大关，其每年新增的包（资产）数量正与 Docker Hub 并驾齐驱。然而，这些模型代表的是一种全然不同且更高级别的制品类型（Artifact）——传统的包治理体系在设计之初，根本无力应对此类新型资产带来的挑战。

下载报告

2025

2024

并非所有漏洞都如表面般危险

全球 CVE（常见漏洞与披露）的公开总量暴增了 20%，已突破 48,000 个，这使得仅凭漏洞数量来进行优先级排查（Triage）变得毫无可能。然而，单纯统计漏洞的原始数量完全释放了错误的信号。在对 248 个备受关注的高危 CVE 进行深度审查后，JFrog 发现：其中真正具备可利用性（Exploitable）的漏洞仅占 11.9%。这意味着，高达 88% 的高危告警其实都只是无效噪音，除了白白浪费开发人员的时间外毫无意义。

下载报告

248

CVE Applicability

Low Applicability
(0% - 20% applicability)

Moderate Applicability
(20% - 80% applicability)

High Applicability
(80% - 100% applicability)

注入漏洞激增 3110%。这得“归功于”你的 AI 编程助手

跨站脚本（XSS）漏洞的 CVE 数量几乎翻了三倍；SQL 注入漏洞的 CVE 数量在短短一年内猛增了 445%；而整个注入类漏洞（Injection Vulnerabilities）更是疯狂激增了 3110%。要知道，这些都是存在了数十年的老牌漏洞类型，技术界早已将其研究透彻且完全可以防范。然而它们如今却卷土重来、疯狂肆虐，其根本原因就在于：AI 辅助开发（AI-assisted development）正在以排山倒海般的体量批量生产这些漏洞，其速度已经彻底碾压了任何人工审查流程。

下载报告

2025

2024

XSS

SQL Injection

Injection