HomeTech NewsCybersecurityWPPConnect WhatsApp Libraries Shipped Secret-Stealing Code

WPPConnect WhatsApp Libraries Shipped Secret-Stealing Code

Two WPPConnect WhatsApp libraries shipped secret-stealing code through the project’s own npm pipeline.

StepSecurity said on 16 November 2026 that npm began serving @wppconnect-team/wppconnect version 2.4.1 at 23:31 GMT on 15 November, and @wppconnect/server version 2.10.82 at 03:09 GMT on 16 November. Both, it said, contain a module that copies the host’s environment variables, .env and configuration files, and WhatsApp session tokens, and uploads them to a Telegram bot. The same research says the code runs when the library is loaded or the server starts, not at install time, so skipping npm install scripts will not stop it.

The npm registry still carries both packages with a deprecation notice on the affected versions: “SECURITY INCIDENT: This version contains code that may transmit host environment, configuration and session data to an external endpoint.” The notices tell users to pin @wppconnect-team/wppconnect 2.4.0 and @wppconnect/server 2.10.81 pending a reviewed release. Registry metadata reviewed for this article shows 2.4.1 published at 23:31 GMT on 15 November and 2.10.82 at 03:09 GMT on 16 November, matching StepSecurity’s times, and shows latest now pointing back to 2.4.0 and 2.10.81.

StepSecurity said 2.4.1 was the latest library version for about 4.5 hours, and that 2.10.82 was never latest. It also said Docker images tagged 2.10.82 and 2.10.83 carried the same code, and that a GitHub nightly build was built from the same commits. After an external researcher opened public issues, it said, the v2.4.1 tag was moved to a look-alike commit without the payload, so the code now shown on GitHub does not match the code that shipped on npm.

How the releases were published is not a finding about who acted

The library release, StepSecurity said, went out through a GitHub Actions trusted-publishing workflow and carries valid SLSA provenance. “The commit behind it had no review.” Provenance, in that account, shows where a package was built, not whether anyone reviewed the source.

Every step, the firm wrote, was taken through a project member account, with the access and release automation the project already had. “GitHub shows which account acted, not who operated it, and we have not established how access was obtained.” That gap matters. The evidence does not identify the maintainer as the person who inserted the code, and it does not establish that an account was taken over. TSN is not filling that gap.

A maintainer wrote on both library issue 2968 and server issue 2734 at 14:44 GMT on 16 November that they had downloaded the npm artifacts without executing them. At that check, they wrote, latest was 2.4.0 and 2.10.81; 2.4.1 and 2.10.82 were deprecated and contained the startup markers identified in the issues; and the pinned versions did not. The maintainer is keeping the issues open because, as they wrote, the registry rollback and source removal leave publishing-account containment, revocation of the destination credentials identified in the issue, and notification to affected users unconfirmed. Users who ran an affected version, they added, should rotate credentials accessible to that process and invalidate exposed sessions.

StepSecurity, reading the project’s Actions logs, says the pipeline also ran for 2.4.3 and 2.10.83, and that those versions are not listed on npm. The maintainer wrote that 2.4.2 and 2.4.3 have no entries in the current registry metadata, and that this does not establish whether they were previously published. That is StepSecurity’s reading of the logs, not a second official package listing.

This is a different fact pattern from Optimizely’s confirmation that Web Experimentation was taken over, from JFrog’s account of an npm worm that armed GitHub repositories, and from earlier npm incidents TSN has covered, including a Tensorlake package incident. Those stories involved compromised or impersonated publishing paths. Here, the research that is in hand says the bad versions left through the project’s own trusted-publishing workflow, and it leaves open how that access was obtained.

What users should check

StepSecurity’s advice is to search lockfiles and node_modules for @wppconnect-team/wppconnect 2.4.1 and 2.4.3, and for @wppconnect/server 2.10.82 and 2.10.83, and to compare Docker image digests rather than tags. If a project loaded one of those versions, or ran an image tagged 2.10.82 or 2.10.83, “treat every secret reachable from that process as stolen, and sign out the WhatsApp sessions.” The firm also suggests a release-age cooldown for newly published package versions. The maintainer’s advice is in the same direction: rotate credentials that the process could reach, and invalidate exposed sessions.

A floor of 155 is not a count of compromised hosts

npm recorded 8,265 downloads of @wppconnect-team/wppconnect in the week from 10 to 16 November 2026. That is a measure of how widely the library is pulled, not of how many machines ran a malicious version, and not of how many secrets left any host. The server package recorded 1,900 downloads in the same week. Neither figure is an impact total.

The only download number StepSecurity offers for the bad library release itself is narrower, and the firm is careful with it. npm, it said, recorded 155 downloads of 2.4.1, but for a window ending 15 November that covers only about 28 minutes of the period in which that version was latest. “155 is a floor, not a count of affected systems.” That wording is StepSecurity’s, and TSN is adopting it as a limit, not as a casualty list. A floor of 155 says only that at least that many registry pulls were counted in an incomplete window. It does not say how many of those installs were actually loaded, how many processes started, or whether any WhatsApp session was used by someone other than its owner. It also says nothing about Docker pulls, nightly builds, or exact-version installs of 2.10.82 after latest had already moved.

The other way of over-reading the evidence is to treat valid trusted-publishing and SLSA provenance as a clean bill of health. In this case they describe a successful pipeline, not a reviewed change. The research leaves the access question open, the maintainer has not claimed that publishing credentials are contained, and there is still no public count of affected users. Until those gaps close, the useful facts are the version numbers, the pin targets, and the advice to rotate anything a loaded process could reach.

Sources

Related stories

Optimizely Confirms Web Experimentation Was Hijacked is the closest recent TSN account of malicious code reaching users through a trusted software channel, and is the next read.

Follow @tsnmediahub on X for breaking AI news as it happens.

Follow on X

Share this story

Latest stories

More in this category

Latest stories

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.