CVE Index中文
L4 · Septic Tank BlewOptimizely · JavaScript · supply chain · skimmer · iOSoptimizely-project-script-injection

Report URI finds malicious Optimizely scripts at 178 organizations

Report URI found malicious Project JavaScript in 188 Optimizely Web snippets used by 178 organizations. The code replaced crypto addresses, targeted card fields and delivered iOS exploit loaders. The vendor confirmed exploitation of an Experimentation account signup flaw and says it removed injected code. Total visitor impact remains unknown.

Disclosed
Nov 16, 2026
Updated Nov 18, 2026
Severity
Septic Tank Blew
No CVE: a breach, not a bug
Splash zone
Visitors to pages that loaded a modified Optimizely Web snippet, including customer-hosted copies of the snippet.
Optimizely · JavaScript · supply chain · skimmer · iOS
In the wild
Yes
Report URI captured malicious production snippets and traced related wallet transfers.

What happened

Report URI found injected code in the Project JavaScript portion of Optimizely Web snippets. That code runs on pages that include the snippet, whether it is loaded from the Optimizely CDN or a customer host. The team observed remote, encoded and inline loaders. Optimizely confirmed exploitation of an Experimentation vulnerability and told BleepingComputer it was in the account signup flow. Its November 17 preliminary report says the signup flaw let new accounts impersonate existing administrator users using public information and repeated email guesses, without stealing or using their passwords. The vendor says no Optimizely administrator account was breached.

The payload could swap cryptocurrency destinations and send page screenshots and cookies readable by JavaScript to the attacker. Card capture targeted values accessible to the page, such as first-party card fields. Report URI says the code could not read card fields isolated in a cross-origin payment iframe. On iPhones and iPads, the loaders used operating-system-dependent paths. Report URI identified Coruna and DarkSword and observed only fingerprinting in the iOS 26 path it examined. In a follow-up, Hudson Rock reports that Ohey retrieved a later DarkSword stage containing code to upload a Keychain database and other files. A request sent with an iOS 26.3 user agent reached another loader and worker, but the chain’s own log marked the route as nonfunctional. Neither finding establishes that a visitor’s device was compromised.

Report URI’s 19:24–19:59 UTC sweep found 186 affected snippets among 1,611 checked. Its full investigation identified 188 altered snippets across 178 organizations. These are sample counts and do not establish how many visitors lost data or assets. The researchers also observed transfers to wallets associated with the payload. Those transfers do not establish total visitor losses or the volume of any card or cookie theft. Their 21:52 UTC recheck found one remaining altered snippet in the measured set. In its November 17 update, Report URI said all tracked affected files were clean by 23:11 UTC on November 16. Its November 17 checks also found that 1688360[.]cc stopped resolving after a registrar hold, webshell[.]lol went on hold, eil786npto9u5bq[.]icu no longer resolved to an IP address and 5vobauua3ojpklx[.]icu still resolved but did not respond. These observations do not resolve exposure before cleanup or verify customer-hosted copies.

On November 17, Bitquery estimated about $2,260 across 19 likely victim payments on November 16, or about $2,210 excluding one ambiguous transfer. It traced much of the roughly $7,040 in earlier observed wallet inflows to a swap rather than direct payments from visitors. These are estimates based on the timing and paths of transfers to the eight known addresses. They do not establish each sender’s intent or account for card theft or losses outside the sampled wallets.

Hudson Rock recovered a malicious minute-ly response in addition to the Optimizely scripts. Ohey’s archived captures show poisoned Kissmetrics and StorifyMe responses on November 13, before the Optimizely captures. Those observations do not show continuous delivery between dates. All three loaded the earlier webshell[.]lol injector. The research confirms the served code and leaves the entry methods for those three providers undetermined. Optimizely’s containment statement applies to its own service.

Optimizely says it patched the vulnerability, removed injected code from affected projects and ended active sessions, revoked tokens and reset access for affected Optimizely accounts, according to its statement to BleepingComputer. Its status is monitoring. Downstream site operators still need to assess and address exposure of their own users’ sessions. Verify deployed copies and caches, inspect logs for the pre-cleanup period, and review session and payment exposure where applicable.

The vendor’s preliminary timeline puts first account creation at 00:37 UTC and first malicious project code at 07:51 UTC on November 16. It records detection at 13:34, blocking at Cloudflare at 16:06, the signup fix at 16:10, session/token revocation at 16:53 and cleanup completion at 22:24 UTC. The 07:51 first-write claim is later than Report URI’s 04:21 captured injection; the reviewed sources do not reconcile that difference. Retain both and request a customer-specific window rather than narrowing log review to the vendor window alone. Cleanup completion and the later independent clean-response check are different observations.

Optimizely excludes Feature Experimentation, SDKs and legacy Full Stack from the identified scope, but says the impersonated users could access whatever Experimentation data their privileges allowed. It says customer passwords were not exposed, does not require general password/API-token resets and permits cleaned Web Experimentation snippets to be re-enabled. Its report says some original Project JavaScript was replaced, so code restoration can require Support. These vendor statements do not erase potential downstream browser exposure, establish visitor loss totals or settle the entry methods at the other script providers. Mandiant-assisted forensics, final root-cause analysis and the final indicator list remain pending.

Crime-scene timeline

  1. Nov 13, 2026

    Archived responses show related scripts before the Optimizely wave

    Later analysis identified poisoned Kissmetrics and StorifyMe responses from November 13. These captures establish when those responses were served, rather than when access began.

  2. Nov 16, 04:21 UTC

    Researchers record an altered Optimizely snippet

    Report URI’s earliest captured injection on an Optimizely snippet loaded the earlier webshell.lol injector on the BBC site.

  3. Nov 16, 08:28 UTC

    The later wallet-replacement loader is observed

    Report URI saw the new payload on the BBC site. Researchers observed card skimming code in a later build at 13:43 UTC.

  4. Nov 16, 21:52 UTC

    One of the 188 known affected snippets still contains the code

    Report URI found one remaining affected snippet in its recheck. The result covers the measured set at that time.

  5. Nov 16, 22:47 UTC

    Report URI publishes its technical investigation

    The researchers describe the payload, affected snippets and containment checks. How the attacker gained project write access remains unresolved.

  6. Nov 16, 23:11 UTC

    The last known altered Optimizely snippet tests clean

    Report URI said in its November 17 update that every affected file it tracked returned clean content by this check.

  7. Nov 16, 2026

    Hudson Rock documents related script delivery through other providers

    Hudson Rock and Ohey verified malicious responses from minute-ly, Kissmetrics and StorifyMe that loaded the same earlier injector.

  8. Nov 17, 01:02 UTC

    Optimizely confirms patch and code removal

    The vendor says it took steps to secure affected accounts and moved to monitoring.

  9. Nov 17, 06:30 UTC

    Researchers publish follow-up evidence

    Hudson Rock and Ohey document earlier captures and later mobile stages. The follow-up also reports Optimizely’s statement identifying the account signup flow.

  10. Nov 17, 07:00 UTC

    Bitquery estimates likely victim payments

    Its on-chain review separates likely victim payments from observed wallet inflows tied to a swap.

  11. Nov 17, 2026

    Optimizely publishes its preliminary customer investigation report

    The vendor describes signup-flow impersonation, a 16:10 UTC fix and 16:53 UTC session/token revocation on November 16. It gives 22:24 UTC as cleanup completion, publishes additional indicators and withdraws the general disable-snippet advice. Final forensics remain pending.

Who said what

We list each claim and who made it. Primary sources (the vendor, the researcher, the registry) are marked. Where reports disagree, we grade by the worst credible case.