On 16 November 2026, malicious JavaScript reached visitors of 178 organisations through the Optimizely Web experimentation platform. Report URI's Scott Helme lists Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax and Ledger among them. The sites' own code was not breached. They loaded a script from a host they already trusted, and that script had been changed.

What the attacker actually changed

Optimizely Web has each customer load a snippet from the Optimizely CDN, for example https://cdn.optimizely.com/js/6395285578973184.js. The snippet includes an optional projectJS field: custom code that runs on every page. The attacker wrote into that field on customer projects and left the rest of each snippet alone.

The file was still served from cdn.optimizely.com with Optimizely's certificate. Report URI says this was not CDN cache poisoning: the CDN served what it was given. How the attacker got write access across unrelated customer projects is, per the post, unknown and "for Optimizely to explain". Report URI says it had no response from Optimizely at publication time.

Three loader variants were seen:

  • A plain <script> tag pointing at https://1688360.cc/track.js (most snippets).
  • The same URL hidden in base64 behind atob(), so grepping for 1688360.cc misses it. Smartsheet was cleaned and then re-infected with this version.
  • A full inline copy of the payload that runs through eval if the domain is blocked.

Customers who serve the snippet from their own domain, such as www.vodafone.com.au/js/<id>.js, were also affected, because the content comes from the same project.

What track.js does

Every visitor got the same script, with all features active. It was rebuilt 26 times during the day.

  • Crypto clipper: swaps valid BTC, ETH/EVM, TRON, Solana, TON, Litecoin, Bitcoin Cash and Dogecoin addresses with the attacker's, in page text, forms, the clipboard and outgoing fetch/XHR. About $7,040 had arrived in the attacker's wallets by 19:25 UTC.
  • Card skimmer: added at 13:43 UTC. It collects a Luhn-valid number, expiry and CVV from forms and POSTs, and sends them to 1688360.cc/card.php. Hosted payment iframes (Stripe, Adyen and similar) were out of reach. Cards typed into a site's own fields were not.
  • Cookie and screenshot theft: sends the URL, non-HttpOnly cookies and an html2canvas screenshot to 1688360.cc/hit.php.
  • iOS exploit kits:
iOS version Chain
13 to 17 Coruna
18.x DarkSword
26 Separate chain; only fingerprints the device

Per the post, iOS 18.7.3 and later, and 26.3 and later, are patched. Pages on 18.6 or earlier reload every 60 seconds to retry the exploit.

Timeline

The first injection was seen at 04:21 UTC. Zoom's snippet was modified at 11:20. Mass clean-up ran from roughly 20:40 to 21:10. At 21:52 UTC, 1 of 188 infected snippets (TalkTalk's) still carried the code.

Which CSPs held

Report URI crawled 256 affected organisations' hosts and tested their CSPs against what the payload needs.

Result Organisations
No CSP at all 68
CSP allowed payload and data theft 50
Payload ran, connect-src blocked data theft (address swapping still worked) 17
Payload blocked 24 (15%)

Allowing cdn.optimizely.com could never help, since the code came from that host. The 24 protected sites had a restrictive script-src listing only needed hosts. Two settings let the payload through:

  • 'strict-dynamic' trusts anything the Optimizely snippet inserts.
  • 'unsafe-eval', which the post says Optimizely's own documentation tells customers to allow. At HSBC, track.js and the data theft were blocked, but the inline copy still ran through eval.

The connect-src result is worth noting. It stopped card, cookie and screenshot exfiltration, but not address swapping, which happens inside the page.

What I would do today

The post's own checklist, in order:

  1. Open every Optimizely project's Project JavaScript and every experiment's custom code. Look for 1688360.cc, atob( and anything you did not write.
  2. Review users, API tokens and integrations with write access, and rotate credentials.
  3. Treat cards entered in your own form fields on 16 November as potentially exposed.
  4. Treat session tokens in non-HttpOnly cookies as exposed and rotate them.
  5. Deploy a CSP with a tight script-src and connect-src, avoid 'unsafe-eval' where you can, and turn on reporting.
  6. Block the attacker domains, including subdomains: 1688360.cc, 5vobauua3ojpklx.icu, eil786npto9u5bq.icu, l7bcsnwc7nzr3uo.icu, webshell.lol, linkforge.cc.

The catch is that a Report-Only CSP only detects. Report URI notes it would have flagged the new script and the new data destinations without breaking anything, but blocking needs enforcement mode. Enforcing a strict script-src against a tag manager or experimentation tool means you have to know what it legitimately loads first. That is work most teams have skipped, and this incident shows what skipping it costs.

Note the source is Report URI, a vendor that sells monitoring for exactly this problem. The timeline, hashes and per-site CSP numbers are theirs and not independently verified here, and the count of 178 organisations is a floor, since they could only check snippets they found.

Sources: