Optimizely compromised, served malware on Zoom, GitLab
On 16 November 2026, malicious JavaScript was served to visitors of 178 organisations through the Optimizely Web experimentation platform. The list includes some of the most visited sites in the world: Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock.
The malware was a multi-purpose kit. It swapped cryptocurrency wallet addresses for the attacker’s, skimmed payment cards, stole cookies and screenshots, and on iPhones and iPads it delivered the Coruna and DarkSword iOS exploit kits. Crypto has already been stolen, with verified transactions to the attackers.
This is exactly the kind of attack Report URI is designed to detect. Our platform can identify unexpected changes to trusted JavaScript, detect new scripts being loaded and reveal suspicious connections to external destinations. In this case, those signals would have exposed the malicious activity, even though the original script was being served from a trusted Optimizely domain. With an appropriate CSP in enforcement mode, organisations could have gone a step further and blocked key parts of the attack before they succeeded.
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.
Most of the malicious code has now been removed. At 21:52 UTC we found it in only 1 of the 188 infected snippets, down from 169 at 20:56 UTC. If your site uses Optimizely Web, or you visited any of the sites below today, read on.
Optimizely Web works by having each customer load a JavaScript “snippet” from Optimizely’s CDN, for example https://cdn.optimizely.com/js/6395285578973184.js. The snippet contains Optimizely’s code, the customer’s experiments and an optional field called Project JavaScript (projectJS): custom code the customer writes in their Optimizely dashboard, which runs on every page.
The attacker wrote their own code into the Project JavaScript of customer projects, and the rest of each snippet was untouched. The file was still served from cdn.optimizely.com, with Optimizely’s certificate, from a host that every affected site already trusted. Here is what was added to Zoom’s snippet:
We saw three versions of the injected loader:
Some customers serve their snippet from their own domain, such as opt.scheels.com/js/
We don’t know how the attacker got write access to so many unrelated customer projects: a compromised Optimizely system, a compromised credential with cross-customer access, or something else. That’s for Optimizely to explain. To be clear, Optimizely’s CDN served exactly what it was given; this was not a CDN cache-poisoning attack. Optimizely’s own sites were affected as well: the snippets for www.optimizely.com and login.optimizely.com both carried the code.
The payload was rebuilt 26 times during the day. The attacker added features, and added wallets but never removed any.
Every visitor received the same script, track.js, and every part of it was active at once. What a visitor experienced depended on their device and what they did on the page.
The script finds any valid wallet address (BTC, ETH and EVM chains, TRON, Solana, TON, Litecoin, Bitcoin Cash, Dogecoin) and replaces it with one of the attacker’s. It validates checksums so it only swaps real addresses. It swaps them in:
It re-scans the page whenever it changes. Copy an address from any affected page, or from anywhere else while that tab is open, and you may paste the attacker’s.
Added at 13:43 UTC. The script watches form submissions and every fetch/XHR POST for a card number (Luhn-checked), expiry date and CVV. When it has all three, it sends them to https://1688360.cc/card.php with navigator.sendBeacon. Cards entered in hosted payment iframes (Stripe, Adyen and similar) were out of its reach. Cards typed into a site’s own form fields were not.
When it swaps an address next to wallet or payment wording, it sends the following to https://1688360.cc/hit.php:
On iPhones and iPads the script loads boot.js from 5vobauua3ojpklx.icu and eil786npto9u5bq.icu. It reads the iOS version and loads an exploit chain in a hidden iframe:
The Coruna loader is byte-identical to a sample we analysed in October from a different campaign. The kit’s own comments say iOS 18.7.3 and later, and 26.3 and later, are patched. Pages on iOS 18.6 or earlier reload every 60 seconds to retry the exploit, which is a visible symptom.
As of 19:25 UTC on 16 November, the attacker’s wallets had received about $7,040, all within roughly nine hours on the 16th:
Funds were swept out at 13:03 to 13:05 UTC. The USDC on Ethereum went to a HitBTC deposit address. The TRON wallet has been in use since March 2026, so this actor was active long before this campaign. We cannot see how many cards or cookies reached 1688360.cc.
We fetched 1,611 distinct Optimizely snippets, found mostly through our own crawl data, and checked each one for the malicious code. 188 snippets belonging to 178 organisations were infected. We kept a copy of every infected file, along with its response headers, SHA-256 hash and fetch time as evidence. The real number is likely higher as we could only check snippets we could find in time.
Ranked by Tranco position within each sector and the status at 21:52 UTC: every snippet except TalkTalk’s had been cleaned.
The earlier webshell.lol wave, through Kissmetrics, StorifyMe and Minute Media, affected other sites and isn’t included above.
Every one of these needs only a CSP in Report-Only mode, so nothing on the site breaks:
It’s possible to go even further and deploy a CSP that could actively block these attacks and prevent any data breach happening. The websites affected by this attack weren’t necessarily running vulnerable code of their own, they were loading JavaScript from a provider they trusted, and that trust was exploited. This is why monitoring what runs in your visitors’ browsers is so important. Report URI gives organisations the visibility to detect when trusted dependencies change or behave unexpectedly, and the tools to help enforce policies that prevent malicious scripts and connections from succeeding.
We crawled 256 of the affected organisations’ own hosts and evaluated their CSP against what this payload needs:
Allowing cdn.optimizely.com was never going to help. The malicious code arrived from the trusted host itself. What separated the 24 protected sites from the rest was a restrictive script-src. A site that lists only the hosts it needs blocks 1688360.cc. Two common settings let it straight through:
connect-src matters. The 17 “partial” sites blocked card, cookie and screenshot theft because they restrict where the page can send data. That doesn’t stop address swapping, which happens inside the page.
If your site uses Optimizely Web:
If you visited an affected site on 16 November:
All of these domains are attacker-owned; block them including subdomains.
Attacker wallets: BTC 19g9rZcPeiz8rUvyWpTRyXXhPnjsWrc3kWETH/EVM 0xC4026b06D82b0F0B363eda0642D61CfbDc17884eTRON THvs6p56XkUiMkJeSc51cAP64u2RsTYqxaSOL FVtL8NMYS9PA1v3HE67XUGCtGA9JaZUvqav2fU2LrV5nTON UQAgM1VqSZQiJLSG5w-ovurq5CZdxLEK3_7CMmwALTA5e09pLTC ltc1q3wp7g6ds0ggg6fl6kncpa4sepn0q3qfacgred5BCH qzkaaf8n6grmpr0d9lx8xutn3r9lk722f566hhpe58DOGE DBEGrxRF7KMSuhTS4c1MFviStU6efdMnu5
We keep an archive of every infected snippet with hashes and timestamps, and can share it with affected organisations and CERTs on request. During research and investigation, once we were confident that Optimizely was the source, we began efforts to reach out via various channels and provide the full details. At the time of publication, we have not had any response.
There are several updates to provide on this incident:
Cryptocurrency transactions were recorded against the attacker’s wallets:
The attacker emptied the wallets early on 17 Nov:
Note that these are on-chain receipts for crypto transactions only. We have no way to see how many credit cards were skimmed, how many cookies or screenshots were stolen, or how many iOS devices were infected before the attack was stopped.
For any third-party script on your website, you can inspect the code it contains right now, but do you know what code it contained last week?
For almost every site on the Web, nobody can answer that question. The URL to fetch the script is the same, but each
If your shop sends customers to a payment gateway to pay, or takes cash on delivery, there’s a comforting thought that goes with it: card numbers never touch your website. The customer types them into Stripe, PayFast, Razorpay, or PhonePe, on someone else’s page. There’
Ten weeks ago, Confiant published research into a Magecart campaign they named HexMage, uncovering a card skimmer that used Ethereum smart contracts to hide its delivery infrastructure.
We’ve been tracking the same campaign independently, and our latest findings show that it’s considerably larger than the original
Most ClickFix pages we look at have been minified, obfuscated and stripped of anything human, but the one we’re about to talk about hadn’t. It was a fake Cloudflare check sitting on foxnevs.com, one letter off Fox News, and it still carried the author’
Until today, the Report URI API and our MCP server were only available on the Business plan and above. As of today, they’re included on every paid plan, at no extra cost. If you’re already on a paid plan, there’s nothing to upgrade and
Over the last few days, a serious security vulnerability in a technology most people have never heard of has caused disruption across the internet.
Government websites have been taken offline, security teams have been working around the clock, and organisations around the world have been urgently reviewing their exposure.
The
Two critical PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619, leaked before PHP could ship a fix. Here’s what’s known, what isn’t, and the three things we did to protect Report URI in the hours before the patch, all of which you can
A macOS ClickFix kit we pulled apart skips all of its own bot and admin checks if you add ?force_fake_captcha=1 to the URL. I guess even malware authors need to test their code too!
Debug Mode
Every developer has written something like this at some point:
if