high

CVE

CVE-2026-55663

CWE

CWE-451, CWE-506, CWE-345

Affected Surface

  • Browsers, users, and internal workflows that opened mirrored HTML for the 24 malicious npm packages reported by OX Security, including live samples such as `ndmxchdjxn2@1.0.0` and `mbxcnsuwgs1@1.0.0` on UNPKG-like package mirrors
  • npm package `mediasoup` `>=3.20.0,<3.20.6` and Rust crate `mediasoup` `>=0.22.0,<0.22.5` when applications used `PlainTransport` or `PipeTransport` with SCTP enabled outside DTLS-protected `WebRtcTransport`

Welcome to Corgea’s weekly briefing. The briefing covers the most important security findings and research from the week.

This edition covers research published from Wednesday, 26 August through Tuesday, 1 September 2026, excluding items already covered in the 25 August briefing and the interim 31 August briefing.

Top Article

24 npm packages turned mirrors into phishing hosts

OX Security deserves first public reporting credit for the 27 August npm mirror campaign. It leads this briefing because it changes the usual supply-chain question. No install hook, import path, or build step has to fire. The package is just the carrier. Once index.html lands on UNPKG or another mirror, the compromise path becomes “someone opened the mirrored page in a browser.” The later mbxcnsuwgs1 wave made that worse by moving the redirect target off package and into encrypted content fetched from api.keyval.org, which let the operator change destinations without republishing the npm artifact.

That makes it a useful companion to the RedC2 import-time cluster, the StubMaker RubyGems typosquat campaign, the @7nohe/openapi-react-query-codegen worm path, and the earlier AsyncAPI require-time compromise. Those stories weaponized install, import, or release workflows. This one turned package infrastructure itself into phishing hosting.

More news

The public advisory for CVE-2026-55663 does not name an outside finder, so the clearest public credit trail stays with the mediasoup maintainers, with ibc authoring the fix. The issue matters if you use mediasoup outside WebRTC’s DTLS boundary. PlainTransport and PipeTransport with SCTP enabled trusted fixed cookie markers instead of a secret-keyed MAC, which meant an on-path attacker could forge COOKIE-ECHO, establish an unauthorized association, and inject DataChannel traffic as a trusted peer.

Read it alongside CVE-2025-62593 in Ray, CVE-2025-34291 in Langflow, and CVE-2026-53362 in the Linux kernel. The products are different, but the same review question keeps coming back: when a system trusts an internal control plane, protocol handshake, or local-only boundary, what proof does it actually require before it accepts traffic as real?

Other news:

From research to remediation

Check whether this pattern exists in your codebase

Turn this research into a remediation workflow. Scan dependencies and package manifests for similar supply-chain risk, then prioritize fixes with reachability context.

Explore dependency scanning Get Demo

References