403 versions in 12 months, one review
New versions per month that fall inside the semver ranges declared in jest@30.5.2's dependency tree (248 packages), Nov 2025 to Oct 2026. Source: npm registry, queried Nov 1, 2026.
Cyber supply chain risk management (C-SCRM) manages the cybersecurity risk in the products, services, and code an organization acquires.
Also known as CSCRM, Cybersecurity Supply Chain Risk Management, ICT SCRM, NIST SP 800-161, SP 800-161
Cyber supply chain risk management (C-SCRM) is the practice of identifying, assessing, and mitigating cybersecurity risk in the products, services, and code an organization acquires. NIST, which calls it cybersecurity supply chain risk management, defines it in SP 800-161 and structures it around supplier contracts, assessments, and provenance controls. Open source dependencies arrive with none of those.
A C-SCRM program works by making suppliers legible. You list them, rank them by criticality, write security requirements into their contracts, assess them before you buy, monitor them for the life of the relationship, and bring them into incident response. Every step assumes a supplier who signed something. On November 1, 2026, we installed jest@30.5.2 into an empty project. The lockfile resolved 248 open source packages, maintained by 116 npm accounts. We have a contract with none of them. In the previous 12 months, those packages published 403 new versions that a fresh install of the same tree would accept without review.
Our position: for open source, C-SCRM has to move its unit of assessment from the supplier to the artifact. Supplier assessments and provenance attestations still have a job. Neither would have stopped the compromise of the npm account Qix in September 2025, which shipped malware in debug and chalk through a legitimate maintainer's account. The control that fits how open source is delivered is reading what each new version's code does, before it runs.
Cyber supply chain risk management is the discipline of treating everything an organization acquires as a source of cybersecurity risk: hardware, software, firmware, and services. NIST SP 800-161 Revision 1 describes C-SCRM as "a systematic process for managing exposure to cybersecurity risks throughout the supply chain." It names three risks in acquired products: malicious functionality, counterfeits, and vulnerabilities from poor development practices.
SP 800-161 is the core C-SCRM document from the National Institute of Standards and Technology (NIST). The first edition appeared in April 2015. Revision 1 followed on May 5, 2022, and an errata update, SP 800-161r1-upd1, replaced it on November 1, 2024. As of November 1, 2026, NIST's C-SCRM project lists no Revision 2 draft.
SP 800-161 organizes C-SCRM into three levels:
The same idea runs through NIST's other documents. SP 800-53 Revision 5, published on September 23, 2020, established a new supply chain risk management control family: twelve controls, SR-1 through SR-12. The Cybersecurity Framework (CSF) 2.0, released on February 26, 2024, moved C-SCRM into its new Govern function as category GV.SC, with ten outcomes from GV.SC-01 to GV.SC-10.
C-SCRM overlaps with third-party risk management (TPRM) but is not the same thing. TPRM covers every risk a third party brings, including financial, operational, and privacy risk. C-SCRM covers the cybersecurity slice and adds product integrity. TPRM asks whether a vendor might get breached. C-SCRM also asks whether the thing you received is the thing the vendor built.
The NIST C-SCRM document that applies depends on the question you are answering. Federal and regulated buyers often cite several of them at once:
| Document | What it gives you | Use it when |
|---|---|---|
| SP 800-161r1-upd1 (Nov 2024) | The full program: three levels, C-SCRM controls overlaid on SP 800-53, and templates for a strategy, policy, plan, and risk assessment | You are building or auditing an enterprise C-SCRM program |
| SP 800-53 Rev. 5, SR family (Sept 2020) | Twelve controls, SR-1 to SR-12, with assignable parameters | You are writing a system security plan or answering an authorization assessor |
| CSF 2.0, GV.SC (Feb 2024) | Ten outcomes, no prescribed controls | You report supply chain risk to executives or a board |
| SP 1326 (July 2026) | A minimum due diligence method for vetting a supplier before an agreement | Procurement has limited time and needs a defensible yes or no |
| SP 800-218, the SSDF | Practices for producers of software | You are the supplier, and a customer's C-SCRM program is asking about you |
Verdict: use CSF 2.0 GV.SC to say what outcomes you commit to, SP 800-161 to say how you run the program, and the SR controls as the evidence trail for individual systems. If you sell software to these buyers, the NIST SSDF is the vocabulary their software attestation questions come from. None of the five tells you how to assess a dependency whose author you have never met.
C-SCRM is important because most code an organization runs was written by someone outside it, and attackers have learned to compromise that code at its source. One compromised upstream account reaches every downstream organization at once.
The Qix compromise shows the scale. On September 8, 2025, a phishing email posing as an npm two-factor reset took over one maintainer's account. Between 13:12 and 13:16 UTC, the attacker published malicious versions of packages including debug, chalk, and ansi-styles. We reported that the affected packages generally receive 2 to 3 billion downloads per week. The payload hooked browser APIs and rewrote cryptocurrency transactions to attacker addresses.
Attacks of this kind are now routine and fast. During the axios compromise, a malicious dependency, plain-crypto-js@4.2.1, was published at 23:59:12 UTC on March 30, 2026. Our automated detection flagged it at 00:05:41 UTC. On May 11, 2026, the TanStack compromise put 84 malicious artifacts into the @tanstack namespace as one wave of the Mini Shai-Hulud worm. An annual supplier review runs on a calendar. These attacks run on a clock measured in minutes.
A typical JavaScript project has dozens of independent upstream publishers, and most of them are individuals. We measured one common development dependency end to end. Every number below comes from the public npm registry API and can be reproduced:
$ npm init -y >/dev/null
$ npm install --ignore-scripts jest@30.5.2
$ npm ls jest
demo@1.0.0
`-- jest@30.5.2The lockfile has 309 entries. Thirty-four are platform-specific optional binaries, of which npm installs only the one matching your machine. The remaining 269 resolved versions belong to 248 distinct packages. Here is what the registry says about them, as of November 1, 2026:
| Measure | Count |
|---|---|
| Distinct packages in the tree | 248 |
| npm accounts listed as maintainers | 116 |
| Distinct publishers of the resolved versions | 60 |
| Resolved versions whose package has a single maintainer | 131 of 269 |
| Resolved versions with an npm provenance attestation | 39 of 269 |
| Packages with an install script | 1 (@parcel/watcher) |
| Versions published to these packages in the past 12 months | 1,869 |
| Of those, versions inside the tree's declared semver ranges | 403 |
The last row matters most for C-SCRM. A dependency range such as debug: ^4.1.0 is a standing order: accept any future 4.x release from whoever holds publish rights. Those 403 in-range releases come to about eight new versions a week that a fresh, unlocked install would have accepted automatically. A lockfile freezes an existing install. It does not review the next version.
403 versions in 12 months, one review
New versions per month that fall inside the semver ranges declared in jest@30.5.2's dependency tree (248 packages), Nov 2025 to Oct 2026. Source: npm registry, queried Nov 1, 2026.
Figure 1 in text: the number of new in-range versions published each month, from November 2025 through October 2026, was 11, 17, 43, 45, 22, 24, 48, 25, 30, 21, 72, and 45, for a total of 403. September 2026 was the busiest month, with 72. The dashed line marks one annual supplier review. SR-6 leaves the review frequency to each organization. The releases keep their own schedule.
Each resolved version carries its supplier identity in registry metadata. The npm registry records who published every version and whether it carries a provenance attestation:
$ curl -s https://registry.npmjs.org/debug/4.4.3 \
| jq '{publisher: ._npmUser.name, maintainers: [.maintainers[].name], provenance: .dist.attestations.provenance}'
{
"publisher": "qix",
"maintainers": ["qix", "tootallnate"],
"provenance": null
}
$ curl -s https://registry.npmjs.org/@babel%2fparser/7.29.12 \
| jq '{publisher: ._npmUser.name, maintainers: [.maintainers[].name], provenance: .dist.attestations.provenance}'
{
"publisher": "GitHub Actions",
"maintainers": ["hzoo", "existentialism", "nicolo-ribaudo", "jlhwung"],
"provenance": { "predicateType": "https://slsa.dev/provenance/v1" }
}debug@4.4.3 is the release that replaced the malicious debug@4.4.2 from the Qix compromise. The registry still records when the removed version went out, though the version itself is gone:
$ curl -s https://registry.npmjs.org/debug | jq '.time["4.4.2"], (.versions | has("4.4.2"))'
"2025-09-08T13:12:39.973Z"
falseToday's jest tree still declares ranges that admit four of the malicious Qix-compromise versions: debug@4.4.2 (through ^4.1.0 in @babel/core), error-ex@1.3.3, strip-ansi@7.1.1, and ansi-styles@6.2.2. Registry removal is what keeps them out of a fresh install now. During the attack window, the ranges let them through.
SP 800-161 controls break on open source wherever they assume a contract. NIST anticipated this. SP 800-161r1 warns that some risks "may circumvent or escape C-SCRM processes, such as those arising from the use of open source software." Maintainers say the same thing from the other side. Thomas Depierre, an open source maintainer, put it bluntly in his essay I Am Not a Supplier: "There is no supply chain here. Because there is no supplier."
Here is how the controls a buyer's program depends on translate to an npm dependency:
| Control | What it assumes | Closest open source equivalent | What the equivalent misses |
|---|---|---|---|
| GV.SC-04: suppliers are known and prioritized by criticality | A finite list of vendors | Maintainer and publisher lists from registry metadata | Publish rights change without notice, and 116 accounts is too many to review by hand |
| GV.SC-05 and SR-8: requirements and notification agreements in contracts | A contract | None. Open source licenses disclaim warranties | No one is obligated to tell you about a compromise |
| SR-6: supplier assessments and reviews | An organization that can be assessed at a defined frequency | Project health signals such as OpenSSF Scorecard | A well-run project with a phished maintainer passes every check |
| SR-4: provenance | Documented origin of components | npm provenance attestations, often via trusted publishing | Provenance proves which pipeline built a package, not that the pipeline was uncompromised |
| SR-11: component authenticity | Counterfeits that differ from the genuine product | Registry signatures and lockfile integrity hashes | Malicious versions published by a real maintainer are authentic |
| SR-10: inspection of systems or components | Physical or logical inspection to detect tampering | Analysis of each new version's contents before install | Coverage depends on running the check at every entry point, including CI and npx |
Verdict: keep SP 800-161's supplier controls for commercial vendors, where contracts and assessments are real. For open source, the controls that survive translation are SR-10, inspecting components, and SR-4(3), which tells you to validate that a component received "is genuine and has not been altered." Treat every new dependency version as its own acquisition, and inspect it.
Provenance deserves a specific warning. npm trusted publishing replaces long-lived tokens with short-lived OIDC credentials from CI. When npm adopted OIDC trusted publishing in 2025, it also made the CLI publish provenance attestations by default. Thirty-seven of the versions in our jest tree were published this way. The TanStack compromise shows its limit. The malicious versions were published through TanStack's own trusted-publisher binding, after attacker code reached the release workflow. Their provenance would name TanStack's real repository and workflow. That answer to the SR-4 question is correct and useless: the pipeline was legitimate, and the code it ran was not.
C-SCRM for open source dependencies keeps the program structure of SP 800-161 and swaps supplier-level evidence for artifact-level evidence. Each element has a cost and a blind spot.
Inventory. A software bill of materials (SBOM) generated from the lockfile satisfies GV.SC-04's "suppliers are known" for everything the lockfile resolves. It misses code fetched at execution time. During the axios compromise, CI jobs that ran tools through npx resolved the malicious version on the fly, outside any lockfile (our analysis of the blast radius has examples). Inventory every place code is fetched, not only the repositories you build.
Criticality. SP 800-161 ranks suppliers by what they supply. For open source, rank by where the code runs. A transitive dependency of a test runner looks low-criticality until you notice it executes in CI next to registry and cloud credentials. Mini Shai-Hulud payloads harvest credentials from exactly those CI environments. A package with an install script runs during npm install itself, before any test or review.
Acquisition gate. The acquisition event for open source is dependency resolution, not a purchase order. Put the check there. A package firewall intercepts each fetch and blocks versions known to be malicious. Socket Firewall works this way: it runs as a proxy between the package manager and the registry and checks each requested package before it reaches disk. Its blind spot is the first minutes after a new malicious publish, before analysis completes. A cooldown that refuses versions younger than a set age narrows that window, but it also delays urgent security patches by the same amount.
Monitoring. GV.SC-07 asks for supplier risk to be "monitored over the course of the relationship." On npm, the relationship changes whenever publish rights change. Socket's New author and Unstable ownership alerts fire when a collaborator publishes a package for the first time. That is the open source equivalent of a supplier changing hands. Neither alert is built for a maintainer account takeover like Qix, where the attacker can publish as a maintainer who has published before. Catching that case takes analysis of the new code.
Incident response. GV.SC-08 brings suppliers into incident response. Open source maintainers will not be on your call, so plan to learn about compromises from advisories and your own monitoring. A clean lockfile checked after the fact does not prove you were never exposed, because CI jobs, npx runs, and fresh installs during the attack window leave no trace in it. Keep install logs and artifact snapshots, so you can answer "did we run it?" with evidence.
C-SCRM is what federal software assurance falls back on now that the government-wide mandates are gone. On January 23, 2026, OMB memo M-26-05 rescinded M-22-18 and M-23-16, which had required secure development attestations from software producers. The memo makes attestations and SBOMs optional. It also says agencies "shall continue to maintain a complete inventory of software and hardware" and must set assurance requirements from their own risk assessments. That is C-SCRM by another name, though M-26-05 never cites SP 800-161. On July 8, 2026, NIST added SP 1326, a quick-start guide for supplier due diligence, built on SP 800-161r1 content.
Regulated buyers in Europe face the same structure. NIS2 Article 21(2)(d) requires supply chain security covering each entity's relationships with "its direct suppliers or service providers." An npm maintainer three levels down your dependency tree is not a direct supplier in that sense. The EU Cyber Resilience Act closes that gap from the product side: manufacturers answer for the open source components inside what they ship.
Both regimes measure supply chain risk by counting suppliers and grading the relationships. Our jest tree has 116 maintainers, no contracts, and 403 in-range releases in a year. The number that predicts exposure is the count of new versions your systems accept without inspection, and no C-SCRM framework yet asks for it.
Manifest confusion is a mismatch between an npm package's registry manifest and the package.json in its tarball, two copies that tools trust inconsistently.
A software bill of materials (SBOM) is a machine-readable inventory of the components, versions, and dependency relationships inside a piece of software.
Software composition analysis (SCA) inventories the third-party code in an application and checks each component for vulnerabilities, licenses, and malware.