Third-Party Script Ownership After the Optimizely Injection

Dim network room with fiber cables under amber light and a distant figure facing away.

Third-party script ownership changed on 16 November 2026, when an attacker used a flaw in Optimizely Web Experimentation to write malicious JavaScript into customer projects, and the vendor CDN served card-skimming, crypto-address swapping, and iOS exploit-kit code to visitors of at least 178 organizations. Pages treated the file as trusted because it still came from Optimizely’s CDN on the vendor certificate.

Updated November 17, 2026

Key Takeaways

  • Attackers wrote JavaScript into Optimizely projects; the vendor CDN served it.
  • A valid Optimizely certificate did not stop the injected project code.
  • Only 24 organizations had a CSP strict enough to block the attack.
  • Smartsheet was cleaned, then re-infected, indicating access remained.

What does third-party script ownership require now?

The malware was not a swapped platform binary. Help Net Security, citing Scott Helme of Report URI, reported it was planted in the Project JavaScript field, which runs on every page, and kept loading from Optimizely’s CDN. The Hacker News reported the rest of each snippet was left untouched and still served from cdn.optimizely.com with the vendor’s certificate, including on sites that host the snippet under their own domain.

Helme’s team checked 1,611 scripts and found 188 infected files across 178 organizations, and said the count is likely higher. Helme said:

The important lesson here is that trusted JavaScript doesn’t always remain trustworthy. A website can be compromised through a third-party dependency without its own code or infrastructure ever being breached.

Why did a valid vendor certificate fail?

The files still came from Optimizely’s CDN on the vendor certificate, so pages treated them as trusted. Optimizely’s statement to BleepingComputer said an unauthorized party exploited a vulnerability in the Experimentation account signup flow and used it to add malicious JavaScript to some customer projects. It said it blocked that path, deployed a fix, removed the injected code, ended active sessions, revoked tokens, reset access for affected accounts, blocked delivery in Cloudflare, and notified affected customers.

Report URI timed the first injection at 04:21 UTC. The Hacker News reported Smartsheet’s script was cleaned once and then re-infected with a hidden version, which indicates the attacker still had access at that point.

At 21:52 UTC, Report URI still saw one of 188 files malicious. Help Net Security later reported the set clean. See also a patched console that stayed rooted.

Which teams actually control experimentation tags?

The product design is the ownership split. Project JavaScript is an optional field meant for the customer’s own code, and it executes on every page, while the vendor serves the snippet.

In this analysis, experimentation owners and security owners need a shared change record: the signup-flow flaw wrote into a field the customer was supposed to author. Before the notice, Report URI said it did not know how the attacker got write access, and that it was for Optimizely to explain. See also a cyber heist traced to a supplier.

What would CSP and write limits have changed?

Helme’s team published both the snippet census and a crawl for policies that would have blocked the attack. Help Net Security reported these figures.

Allowing Optimizely’s domain did not help, because the malware was delivered from that domain. A setting the vendor documents, which permits eval(), let a backup copy run on HSBC’s site.

Measure Figure Scope
Scripts examined 1,611 Snippet check
Infected files 188 Same check
Organizations 178 Owners of those files
CSP strict enough to block 24 (15%) Affected-site crawl
No CSP 68 Same crawl
Policy let malware run and send data 50 Same crawl

Three loader forms limit inventory. Some called the payload URL directly, some hid it with Base64 and atob(), and some embedded a copy that runs if the domain is blocked. Strict-dynamic or unsafe-eval let that copy run on some sites, Report URI warned.

In this analysis, the checks Report URI and StepSecurity published remove specific false comforts:

  • Review Project JavaScript and every experiment’s custom code for 1688360, atob(, or any script you did not write.
  • Limit users, API tokens, and integrations that can write, then rotate those credentials.
  • Deploy a strict script-src and connect-src. Allowing cdn.optimizely.com does not contain injected project code.
  • Keep third-party tags off checkout, and take cards in a cross-origin payment iframe. The skimmer could not read those fields or HttpOnly cookies.

Frequently asked questions

What changed in the Optimizely injection?

Optimizely told BleepingComputer an unauthorized party used a signup-flow vulnerability to add malicious JavaScript to some customer projects. Report URI said the vendor CDN then served that code to visitors of at least 178 organizations.

Did a valid certificate mean the script was safe?

No. The files still came from Optimizely’s CDN with the vendor certificate, and pages treated them as trusted.

Would allowing Optimizely in a CSP have blocked this?

No. Allowing Optimizely’s domain did not help, because the malware was delivered from that domain. Help Net Security reported only 24 organizations, or 15%, had a policy strict enough to block it.

Did delivery mean the website’s own systems were breached?

Not necessarily. The Hacker News reported Cynet’s finding that serving the code does not mean the site’s own systems were compromised. Hudson Rock said successful device exploitation was not shown.

Notice an error?

Help us improve our content by reporting any issues you find.