Skip to content
NewA calmer, faster ZeroHour

Attackers hit 42 GitHub repos with files too large to search

Original title: Attackers compromised 42 GitHub repositories by using files too large to be searched

CSO Online ()5 min read2 · 1 read
Part of a story covered by 5 sourcestestmetest npm worm armed 42 GitHub repos with oversized payloadsMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

Oversized wrappers and a native addon hid one CI-publisher payload while an npm worm stole tokens and spread through Actions.

JFrog Security Research said attackers compromised 42 third-party GitHub repositories using payload blobs large enough to evade GitHub search and the contents API. Repositories held hooks for AI coding assistants and IDEs, plus polymorphic wrappers as large as 13.7MB that decoded to the same hidden payload. The related npm chain, linked to the testmetest and Miasma-family activity, stole credentials, poisoned CI, and republished malicious packages. JFrog advised treating a full-worm install as a host compromise and rotating exposed credentials.

Unconfirmed: SafeDep noted uncertainty about whether the payload belongs to the Miasma family.

Key facts
  • Wrappers reached 13.7MB, beyond what GitHub code search and the contents API return.
  • A compiled native addon decrypts .napi_cache at runtime rather than carrying its own payload.
  • Malicious npm releases ran index.js via binding.gyp and used Bun 1.2.18 after AES-128-GCM decryption.
  • JFrog said to block 43.228.157.68 and remove unauthorized chore/ci-* and deps/upd-* branches.
  • SafeDep first documented the npm side on October 22 and was uncertain about Miasma lineage.
Security assessment · AI, from the article

Impact: Stolen developer credentials and poisoned repositories and CI publishing

What to do
  1. 1.Treat installation of a full-worm version as a host compromise.
  2. 2.Revoke and rotate GitHub, npm, cloud, SSH, and AI-provider credentials.
  3. 3.Enumerate repository trees directly and do not rely on GitHub code search.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

Indicators of compromise

auto-extracted · verify before use · export allAll
TypeIndicatorContext
ipv443.228.157.68ve fake CodeQL artifacts named scan-output . Block hxxp[:]//43[.]228[.]157[.]68:80 , and related malicious infrastructure at network
urlhttp://43.228.157.68:80and remove fake CodeQL artifacts named scan-output . Block hxxp[:]//43[.]228[.]157[.]68:80 , and related malicious infrastructure at network c

From the original article

5 min read at csoonline.com

A security research firm reported Monday that it had identified breaches of 42 third-party GitHub repositories in which attackers had deployed payload blobs large enough “to evade the GitHub search and API paths normally used to find this kind of compromise.”

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.