Skip to content

Response, New York. Noah Park: Check Point Patches Actively Exploited Management Server Path Traversal. Detection, Dallas. Ava Okello: Report Finds Nearly Half of Deployed Detections Need Attention Before SOCs Can Trust Them.

SOCtember

Always First. Fast SOC News.

StepSecurity illustration for the WPPConnect compromise: a green box labeled GitHub tag v2.4.1 and a purple box with a warning sign labeled npm tarball 2.4.1, separated by a not-equal sign, over the caption Same version. Different code.
Image: StepSecurity (fair use). Cover art from the StepSecurity post "WPPConnect Compromised," published Nov. 16, 2026.

Response

WPPConnect's Own Release Pipeline Shipped a Stealer That Sent Server Secrets and WhatsApp Sessions to a Telegram Bot, With Valid Provenance and a GitHub Tag Rewritten to Hide the Code, StepSecurity Says

Two npm releases, Docker images and a nightly build of the WhatsApp automation libraries carried the payload, and it ran on import, so --ignore-scripts did not help. StepSecurity says anyone who loaded them should treat every reachable secret as stolen.

Noah Park, ResponseNew York6 min read

NEW YORK - Malicious releases of WPPConnect, a set of open-source libraries for automating WhatsApp Web, went out through the project's own trusted-publishing workflow late Sunday and early Monday, carrying a module that uploaded host environment variables, configuration files and WhatsApp session tokens to a Telegram bot, the software supply chain security company StepSecurity said.

The affected packages are @wppconnect-team/wppconnect 2.4.1 and @wppconnect/server 2.10.82, along with 2.4.3 and 2.10.83, which the pipeline built but npm does not list. Docker images of wppconnect/wppconnect-server and the project's GitHub nightly release also carried the code. StepSecurity researcher Owais Shaikh, who checked the npm registry, GitHub's API and Actions logs, the Sigstore transparency log and Docker Hub, said three details change how defenders should respond: the code runs when the library is imported, not when it is installed; the releases carried valid publishing credentials; and the release tag was rewritten after the code was reported.

A clean-looking release

The sequence began at 22:34 UTC on Nov. 15, when two unsigned commits landed on the master branch of the library repository with no pull request. One added a 208-line file, src/utils/diagnostics.ts, and the other imported it into the browser controller. A release workflow was dispatched by hand within a minute, and npm began serving 2.4.1 as the latest version at 23:31 UTC. The server repository received a one-line change at 02:43 UTC on Nov. 16, which automatically triggered the 2.10.82 release. Every step was done through the joaosouz4dev account; StepSecurity said GitHub shows which account acted, not who operated it, and it has not established how access was obtained.

The library's publish workflow had no GitHub environment with required reviewers, so nothing stood between the tag and the registry. npm shows both releases published by GitHub Actions under a trusted-publisher identity, and 2.4.1 carries a SLSA provenance attestation that StepSecurity decoded. Every field was accurate, and the commit it named was the malicious one. "Provenance proves where a package was built, not whether anyone reviewed the source," the company wrote. A trusted-publisher check passes both releases. A provenance check flags only the server release, which has no attestation.

StepSecurity post box headed Check your projects and servers now, listing the affected WPPConnect library and server versions, Docker image windows and the advice to treat reachable secrets as stolen.
Image: StepSecurity (fair use).The response box from the StepSecurity post.

About 30 seconds after an outside researcher opened public issues on both repositories at 04:00 UTC on Nov. 16, both default branches were force-pushed. Until Nov. 17, the v2.4.1 tag pointed to a look-alike commit with the same message and timestamp as the one npm built from, but without the payload, so anyone reviewing the tagged source on GitHub would not have found the code that shipped. The tags for all four versions were deleted at 14:06 UTC on Nov. 17. The original commits are still reachable by hash.

StepSecurity timeline table of the WPPConnect compromise in UTC, from unsigned commits on Nov. 15 at 22:34 to npm packuments being modified around 04:03 to 04:06 on Nov. 16.
Image: StepSecurity (fair use).The incident timeline from the StepSecurity post.

What the module took

The module runs whenever the library is loaded, and the server loads its own copy at startup, so a container running the Docker image ran it on start. The library has no install script, which is why npm install --ignore-scripts offered no protection. It collected the hostname, user, OS, Node version and public IP; every environment variable with a value of 3 to 499 characters, which covers API keys, database URLs and cloud credentials; the first 12,000 characters of .env, .env.local, .env.production, config.js, config.json and ormconfig.json; and up to five session files from the tokens folder, which the project's own type definitions describe as the credentials for a WhatsApp Web session. The report went to the Telegram Bot API, at most once a day per host, with a marker file named .wpp-support-diag or .wpp-diag written to the temp directory.

The server variant waited 30 to 240 seconds and skipped hosts that looked like CI runners or sandboxes. The library variant had no such checks, so a CI job that imported it would have sent its environment, including any secrets passed to the job. npm recorded 155 downloads of 2.4.1, which was the latest version for about 4.5 hours, but StepSecurity said that figure covers only about 28 minutes of the window and is a floor.

GitHub issue 2968 on the wppconnect-team/wppconnect repository titled Security: npm release @wppconnect-team/wppconnect 2.4.1 (still latest) uploads secrets to a Telegram bot, describing what the code does.
Image: GitHub (fair use).The public issue that reported the malicious release, wppconnect-team/wppconnect #2968.

What responders should do

Search lockfiles and node_modules for library 2.4.1 and 2.4.3 and server 2.10.82 and 2.10.83, and check 2.4.2 as a precaution. Compare Docker image digests rather than tags: any wppconnect/wppconnect-server image tagged 2.10.82 or 2.10.83, or pulled between 02:48 and 03:54 UTC on Nov. 16, should be treated as infected. Look for the marker files, and search DNS, proxy and egress logs for api.telegram.org and api.ipify.org from hosts that run WPPConnect, keeping in mind that both are shared services and a hit alone is not proof.

If the code ran, StepSecurity says to rotate, from a clean machine, every environment variable available to the process, every value in those config files and any secret passed to CI jobs that imported the library, then unlink the WhatsApp devices, delete the old session files and pair again only after pinning library 2.4.0 and server 2.10.81 with a committed lockfile and an override, since the server pulls the library in at ^2.3.6. The maintainers said on Nov. 16 that containment of the publishing account and revocation of the Telegram credentials were not yet confirmed, and StepSecurity advised treating any new release as untrusted until its tarball has been reviewed.

For teams that publish with trusted publishing, the lesson is structural: put the publish job behind a GitHub environment with reviewers who did not push the tag, require pull requests and signed commits, block force-pushes and add a release-age delay on the consuming side. A 24-hour delay would have kept 2.4.1 out.

About the Author

Noah Park

Noah Park covers incident response, containment, and blue-team operations for SOCtember from New York.

Related stories

Response desk