A bug in Optimizely's signup flow let an attacker inject malware into 178 companies' websites
An attacker used a flaw in Optimizely's account signup flow to inject malicious code into customer tracking snippets, serving crypto-theft and card-skimming scripts through sites run by CNN, Salesforce, Zoom and dozens of other brands.

An attacker exploited a bug in Optimizely's account signup flow on November 16 to inject malicious JavaScript into customer tracking snippets, serving the code to visitors of sites run by CNN, the BBC, Salesforce, Zoom, HSBC and at least 173 other organizations before Optimizely shut down the vulnerable path.
What happened
Optimizely told BleepingComputer that the attacker exploited "a vulnerability in the Experimentation account signup flow" to write code directly into customers' Project JavaScript, the field marketers use to run A/B tests and personalization logic on their own pages. Because the snippet is served from cdn.optimizely.com, a domain every affected site already trusted, the injected code loaded without tripping most security controls built to watch for unfamiliar scripts.
Security researcher Scott Helme, founder of the monitoring vendor Report URI, first spotted the code in a BBC snippet at 4:21am UTC on November 16. By the time his team finished reviewing 1,611 Optimizely snippet files that day, 188 of them, belonging to 178 organizations, were infected. The payload swapped cryptocurrency wallet addresses for the attacker's own, then, starting at 1:43pm UTC, began watching form submissions for card numbers, expiration dates and CVVs typed directly into a site's own fields; card data handled through hosted Stripe or Adyen iframes was out of reach. On iPhones and iPads running older iOS versions, it loaded exploit kits Helme named Coruna and DarkSword. His tally puts the attacker's take at roughly $7,040 in crypto by 7:25pm UTC, emptied from the wallets early the next morning. Report URI sells the monitoring tools it argues would have caught the activity, worth flagging since the post also markets its own products, though its timeline matches BleepingComputer's reporting and Optimizely's own statement.
Optimizely said it blocked the vulnerable signup path, removed the injected code, ended active sessions, revoked tokens, reset access for affected accounts and notified every customer involved; its status page still listed the incident as "Monitoring" on the morning of November 17. Salesforce, one of the named organizations, told BleepingComputer it had disabled Optimizely's scripts across its preview and production environments as a precaution while it investigates.
Why this is a marketing problem, not only a security one
Optimizely's tag is not server infrastructure that IT quietly patches. It is martech: a script marketing teams request, configure and leave running on every page of a site, often with several people holding write access to whatever experiment is live. That access is exactly what this attack rode in on, and it is the reason the damage reached 178 companies' visitors rather than Optimizely's own systems alone. A tool your team chose and controls, not a server someone else manages, was the way in.
Helme's team also checked whether a Content Security Policy would have stopped the payload, across 159 of the affected organizations' sites. The results show how exposed a typical marketing stack already is: 68 sites had no CSP at all, 50 had a policy loose enough that both the malicious script and the stolen data could leave the site, and 17 blocked outbound data but still let the wallet-address swap happen inside the page. Only 24 sites, about 15%, fully blocked execution.
"Allowing cdn.optimizely.com was never going to help."
Helme wrote, since the malicious code shipped from the very domain a trusting policy was built to allow. He added a line aimed squarely at marketing and comms teams, not just security: consider whether you need to tell your own customers that your site served them malware. Report URI filed the incident under its "Magecart" category, the same label used for years of card-skimming attacks delivered through third-party scripts rather than a retailer's own code. Optimizely's reach, spanning experimentation tags on sites across finance, retail, media and healthcare, is simply what made this particular Magecart-style incident land on 178 companies at once instead of one.
What to check this week
If your site runs Optimizely Web Experimentation, the fix isn't just waiting on a vendor patch note. A few checks are specific enough to do today:
- Search every project's Project JavaScript and every experiment's custom code for the domain "1688360.cc", any use of atob() you didn't write, or scripts from outside your own team.
- Pull the list of who has write access to your Optimizely account, plus API tokens and integrations, and rotate anything you can't immediately account for.
- Check Optimizely's change history log, filtered to November 16, for edits to Project JavaScript nobody on your team made.
- If your own site forms, not a hosted Stripe or Adyen checkout, collect card numbers, treat anything entered on November 16 as potentially exposed and loop in finance.
- Find out whether your site has a Content Security Policy at all, and if it does, whether it restricts connect-src rather than leaning on an allowlisted domain for protection.
Decide with legal and support, not only with IT, whether customers need to be told their visit to your site may have served them malware, and who owns sending that message.
Tools & AnalyticsBloomreach acquires AI-visibility platform AthenaHQ, tying AI citations to revenue
Sienna Mcpherson · 5 min read
AIAI search sends most YouTube software citations to reviewers, not vendors
Sienna Mcpherson · 4 min read
Distribution