Key Takeaways
- Attackers abused an Optimizely flaw to plant malware in snippets used by 178 organizations on November 16.
- The payload swapped crypto addresses and went after card data, cookies, screenshots and unpatched iPhones.
- Allowlisting the vendor’s domain did nothing, since the malicious code arrived from the trusted host itself.
- SecureChain checks packages before the build, because Content Security Policy (CSP) trusts whatever ships in first-party code.
On Monday, November 16, 2026, visitors to the websites of Zoom, the BBC, CNN, HSBC, IKEA and Ledger were served malicious JavaScript. The attackers didn’t need to break into those companies’ servers or trick anyone into shipping a bad release. The code arrived inside a script the sites had included on purpose, from a vendor they trusted, over a valid TLS connection.
Scott Helme and the team at Report URI spotted the attack and tracked it live on X. His full write-up counts 188 infected snippets out of 1,611 checked, belonging to 178 organizations across banking, retail, media, airlines, healthcare and crypto. Treat that number as the lower bound, since Report URI could only check the snippets it could find.
Verify Your npm Packages Before They Ship with SecureChain
The Optimizely attack ran in the browser, but the same crypto-stealing tricks have already shipped inside compromised npm packages. SecureChain for JavaScript rebuilds packages from vetted source and screens them for malware before they reach developers or CI, and it works alongside Artifactory, Nexus, and GitHub Packages.
What happened in the Optimizely supply chain attack
Optimizely Web Experimentation is an A/B testing and personalization platform. Customer sites load a JavaScript snippet from cdn.optimizely.com, or from their own domain in some cases, that carries Optimizely’s own code, the customer’s experiments and an optional field called Project JavaScript (projectJS), which runs on every page.
The attacker wrote a small loader into that field in many customer projects and left the rest of each snippet untouched. Most versions simply pulled track.js from 1688360[.]cc. Some hid the URL in base64 behind atob(), so a plain-text search for the domain came back empty. A third variant carried the whole payload inline and ran it through eval if the external domain was blocked.
Report URI ruled out CDN cache poisoning. Optimizely’s CDN served exactly what it was given, and the snippet on Optimizely’s own login page carried the code too. Optimizely confirmed on its status page that an unauthorized party exploited a vulnerability in Experimentation to add malicious JavaScript to some customer projects. A spokesperson told Cointelegraph the weak point was the Experimentation account signup flow, and that the company had fixed the flaw, removed the injected code, ended active sessions, revoked tokens and reset access for affected accounts.
The first injection showed up in the BBC’s snippet at 04:21 UTC, initially loading a script from webshell[.]lol before track.js took over later that morning. Report URI says every infected file it tracked was clean by 23:11 UTC the same day.
One script doing five jobs
Once track.js took over, every browser that fetched it got the same script, and the inline fallback carried the same code. By late in the day it held all five capabilities at once, and what actually ran depended on the device and on what the visitor did. The crypto clipper did the earliest damage. It hunted for wallet addresses across Bitcoin, Ethereum and other EVM chains, TRON, Solana, TON, Litecoin, Bitcoin Cash and Dogecoin, and it checked their checksums so it only swapped valid addresses. Then it swapped in the attacker’s address across page text, form fields, links, the clipboard and outgoing fetch and XHR requests.
A card skimmer was added at 13:43 UTC. It watched form submissions and outgoing fetch and XHR POST requests for a Luhn-valid card number, expiry date and CVV, then sent all three to the attacker’s server. Hosted payment iframes from providers such as Stripe or Adyen were out of reach; card fields built into a site’s own forms were not.
Whenever the script swapped an address next to wallet or payment wording, it also sent off the page URL, whatever cookies the page could read through document.cookie and a screenshot of the page rendered with html2canvas. iPhone and iPad visitors faced an extra layer. A hidden iframe loaded the Coruna exploit kit on iOS 13 through 17 and DarkSword on iOS 18.x, while a separate iOS 26 chain was only fingerprinting the device when Report URI looked at it. Comments in the exploit code itself say iOS 18.7.3 and later, and iOS 26.3 and later, are patched.
The money trail is small – Report URI counted about $7,040 landing in attacker wallets by 19:25 UTC on November 16 and ties those transfers to the address swapping. StepSecurity, working independently, put the figure at roughly $6,600, according to The Next Web. Neither number measures the incident. Public reporting hasn’t established how many card details, cookies or screenshots were sent out, or whether any iPhone or iPad was actually compromised.
The payload was also rebuilt 26 times during the day as the attacker added features and wallets. A scan run after the cleanup would show a perfectly normal script, even though visitors received something very different hours earlier. Historical script captures, response timestamps and browser telemetry are the evidence that counts here.
Why allowlisting Optimizely did nothing
A green build, a clean malware or vulnerability scan and a valid TLS certificate say very little about what a remotely hosted script will do after deployment. Most scanners check a site’s own code and servers, while this payload lived on Optimizely’s side, changed through the day and was hidden behind obfuscation in some variants. Content Security Policy (CSP) is the main browser-side defense, and Report URI checked the CSP on affected organizations’ own sites. In its breakdown, 68 organizations had no CSP at all and another 50 had a policy that allowed both the payload and the data theft. A further 17 ran the payload but blocked exfiltration through connect-src, which stopped card, cookie and screenshot theft but not the address swapping happening inside the page. Only 24 blocked the payload outright.
Three settings explain most of the failures. Listing cdn.optimizely.com in script-src was never going to help, because the bad code came from that exact host. Under 'strict-dynamic', a snippet trusted by nonce or hash passed that trust to every script it inserted. And 'unsafe-eval', which Optimizely’s documentation asks customers to allow, let the inline fallback run even where track.js itself was blocked.
Subresource Integrity, the usual answer to “what if the CDN changes the file?”, is a poor fit too. An experimentation snippet changes every time someone edits a test, so a pinned hash would block legitimate updates unless someone reviewed and re-pinned every change. OWASP’s third-party JavaScript guidance recommends treating publishing access to these platforms as access to application code and monitoring vendor scripts for changes, which matters most for code that changes by design.
The same trust problem lives inside your own bundle
Strip away the vendor and the core failure is simple. Trust was granted to a name and a location, and whatever arrived from that location ran. A similar decision gets made constantly in a typical JavaScript codebase, once for every package in the dependency tree, most of them transitive and rarely reviewed.
It has already been exploited in almost exactly this way. On September 8, 2025, a phishing email impersonating npm support tricked the maintainer behind chalk, debug and 16 other packages into handing over account access. Together those packages saw more than 2 billion weekly downloads. The malicious versions carried a browser-side crypto clipper that rewrote wallet addresses, the same job track.js was doing on November 16.
One difference makes the package version of this attack harder to catch in the browser. A compromised dependency gets bundled into the application and ships inside the site’s own JavaScript, served from the site’s own domain. There is no unfamiliar host for script-src to block. 'self' lets it straight through, and an integrity hash generated from that build simply vouches for the malicious code inside it. A tight connect-src can still stop stolen data from reaching an unapproved destination, but CSP has no way to tell malicious code from legitimate code once both sit inside an authorized bundle. The best place to stop that kind of attack is before the build, at the point where packages enter the codebase.
Where SecureChain fits
SecureChain doesn’t run in a browser, and it would not have touched an Optimizely snippet. Hosted vendor scripts need CSP, monitoring and tight control over who can publish to them. What SecureChain does is replace blind trust with verification for the packages that become a team’s own code.
Instead of installing straight from the public registry and hoping the latest upload is what the maintainer intended, teams point npm, yarn or pnpm at SecureChain for JavaScript. It works alongside Artifactory, Nexus and GitHub Packages, so existing workflows stay the same. JavaScript is the obvious fit for this story, but SecureChain also covers Python, Java, Go and PHP, all currently supported. Every package is rebuilt from vetted source in hardened, isolated environments and screened for known malware and suspicious code patterns before it reaches a developer or a CI pipeline. Each build is signed and carries SLSA Level 3 provenance, with VEX data and, where available, an SBOM. Together they show how a package was built, what’s inside it and which known vulnerabilities actually affect it. Known vulnerabilities are patched under SLA, and when upstream maintenance ends, Endless Lifecycle Support keeps fixes coming for as long as the package stays in the stack.
For hosted scripts, the job is watching what arrives in the browser. For packages, it’s verifying what arrives in the build before it turns into first-party code that browser controls treat as the site’s own.
What security teams should check now
Teams running Optimizely Web should open every project’s Project JavaScript and every experiment’s custom code and search for 1688360, webshell, atob( and anything nobody remembers writing. Review which users, API tokens and integrations can write to the account, and rotate credentials. If card fields live in the site’s own forms, treat cards entered between 13:43 and 23:11 UTC on November 16, while that site’s snippet was infected, as potentially exposed, treat secrets in non-HttpOnly cookies the same way and bring in the incident response and legal teams.
Everyone else should block 1688360[.]cc, webshell[.]lol, 5vobauua3ojpklx[.]icu, eil786npto9u5bq[.]icu, l7bcsnwc7nzr3uo[.]icu and the related hosts in Report URI’s indicators table, including subdomains. The longer-term project is an inventory of every script that runs in visitors’ browsers, each with a named owner. Pair it with a CSP tested in report-only mode before enforcement, with a tight script-src, a restrictive connect-src and no 'unsafe-eval' unless something truly needs it.
On the package side, commit lockfiles, pin versions for anything that touches money or credentials, give new releases a short cooldown before they reach production and pull from a source that verifies what it serves.
iPhone and iPad users who visited an affected site that day should update iOS or iPadOS. Anyone who sent crypto on November 16 with an affected site open in a browser tab should check the destination against the intended address, and anyone who paid by card on an affected site should keep an eye on their statements.
The question still hanging over this one
Scott Helme asked publicly whether visitors should expect to hear from the organizations involved. Optimizely says it has notified its affected customers. Whether those customers tell the people who loaded their pages that day is still an open question. For everyone else, the more useful test is whether anyone on the team would notice if an approved script or package changed tomorrow.