who we are

security expertise, democratized

A close-knit team of experienced security professionals with a shared conviction: real security expertise should be accessible to every organization, not just those with enterprise budgets.

OPTIMJACK: Optimizely Hack Turns Major Websites Into Malware Delivery Points
← back to insights
InsomniSec Team · November 17, 2026

On Monday, November 16, attackers used a flaw in Optimizely, a widely used website testing and personalisation service, to slip malicious code into the websites of at least 178 organisations, among them the BBC, CNN, Fox News, Zoom, Salesforce, GitLab, HSBC, and IKEA. None of those companies were breached directly. The code arrived through a trusted marketing script they had chosen to load on their own pages, so their own security controls let it through.

For roughly 17 hours, visitors to affected sites could have had cryptocurrency payments redirected to the attacker, payment card details stolen from checkout forms, browser cookies and page screenshots taken, and older iPhones and iPads targeted with exploit kits. Researchers traced about $7,000 in cryptocurrency to the attacker's wallets on Monday. Optimizely says it has fixed the flaw, removed the code, and notified all affected customers, but key questions remain open, including how long the flaw existed and how many visitors were exposed. Any organisation running Optimizely Web Experimentation should treat Monday as a possible exposure window: check its scripts, review who can change them, and decide whether customers need to be told.

The campaign has been dubbed OPTIMJACK by researchers. No CVE has been assigned. Optimizely told BleepingComputer the attacker exploited a vulnerability in the Experimentation account signup flow, but has not published technical details of the flaw.


Who is affected

Directly affected: organisations using Optimizely Web Experimentation whose projects were modified. Report URI examined 1,611 Optimizely snippets and found malicious code in 188 of them, belonging to 178 organisations. Because only a sample was examined, the true number is likely higher. Optimizely has confirmed that some Web Experimentation customers were affected but has not said how many.

Named organisations in public reporting include:

  • Technology and SaaS: Zoom, Salesforce, GitLab, Smartsheet, Bitly, ManageBuilding, Stamps.com
  • News and media: BBC, CNN, Fox News, Peacock
  • Financial services: HSBC (including HSBC Insurance), Equifax, Mutual of Omaha, IG, Ledger
  • Retail: IKEA, American Eagle, American Girl, Scheels
  • Travel, telecoms, and non-profit: Air France-KLM, TalkTalk, Change.org, UNICEF USA

Optimizely's own main website and customer login page were also serving the malicious code. BleepingComputer independently confirmed, using urlscan.io scans, that the malicious script loaded or attempted to load on many of these sites.

Salesforce has confirmed it is investigating. It told BleepingComputer that, as a precaution, Optimizely scripts are disabled across its preview and production instances, and that it will notify affected customers if it finds unauthorised access to their data.

Also affected: sites that self-host or proxy the Optimizely snippet from their own domain. The injected code was part of the project configuration, so it travelled with the snippet wherever it was served from.

Indirectly affected: anyone who visited an affected site on Monday, particularly those who entered card details into a site's own checkout form, copied or pasted a cryptocurrency address, or browsed on an iPhone or iPad running an older version of iOS.


How the attack worked

Optimizely Web Experimentation lets marketing and product teams run A/B tests and personalise pages. Each customer's site loads a JavaScript snippet from cdn.optimizely.com, and each snippet includes an optional field called Project JavaScript, where customers can add their own code. That code runs on every page load, for every visitor.

The attacker wrote a small loader into Project JavaScript and left the rest of the snippet untouched. Because the file was still served from Optimizely's own CDN, over Optimizely's certificate, browsers and most security controls had no reason to treat it as suspicious. Sites with a Content Security Policy that allows cdn.optimizely.com were not protected, since the loader came from an allowed source.

The loader appeared in three variants:

  • A direct reference to the attacker's domain
  • A base64-encoded version of the same reference, decoded with atob()
  • A version carrying a full embedded copy of the payload, which runs even if the remote file is blocked or fails to load

That third variant matters. Hudson Rock and others have warned that blocking the attacker's domain is not enough on its own, because some poisoned snippets did not need to reach it.

How the attacker got in

Optimizely confirmed to BleepingComputer that an unauthorised party exploited a vulnerability in the Experimentation account signup flow and used it to add malicious JavaScript to some customer projects. In other words, the weakness was in how new accounts are created, and it gave the attacker a way to write to projects belonging to existing customers. Optimizely has not explained the mechanism, how long the flaw was present, or how many projects were changed.

In its statement, Optimizely said it has:

  • Blocked access to the vulnerable path and deployed a fix
  • Removed the injected code from affected projects
  • Ended active sessions, revoked tokens, and reset access for affected user accounts
  • Blocked any code from being delivered through Cloudflare
  • Notified all affected customers

It says further updates will be published at status.optimizely.com.

One detail suggests the attacker's access outlasted the initial cleanup: Smartsheet's script was re-infected after it had been cleaned, according to The Hacker News. That is a good reason to keep checking your own snippets for some days, rather than treating a single clean result as the end of the matter.


What the malware did

The payload was revised several times through the day, adding capabilities as it went.

  • Crypto wallet clipper (track.js): Watches page text, form fields, links, the clipboard, and outgoing requests for cryptocurrency wallet addresses, and swaps them for attacker-controlled addresses. It only swaps addresses that pass checksum validation and appear near payment-related wording, which makes it quieter and harder to notice in testing.
  • Card skimmer: Added at about 13:43 UTC. Captures card number, expiry date, and CVV from a site's own form fields and sends them to 1688360[.]cc/card.php. Per StepSecurity, it cannot read card details entered into a payment provider's cross-origin iframe, such as Stripe or Adyen hosted fields.
  • Session and page theft: Collects cookies that are not marked HttpOnly, page URLs, and a screenshot of the page, sent after an address swap near payment wording. Visitors who were logged in to sensitive services on an affected site should watch those accounts.
  • iOS exploit delivery: Loads a boot.js script that checks the visitor's iOS version and opens an exploit chain in a hidden iframe:
    • Coruna for iOS 13 to 17.2.1
    • DarkSword for iOS 18.4 to 18.7
    • A separate chain for iOS 26 that, so far, only collects device details

On the iOS question, the evidence points to attempted delivery rather than confirmed compromise. Cynet described what it observed as conditional delivery in an emulated browser, "not a successful compromise" of a physical device, and Hudson Rock said successful device exploitation has not been demonstrated. Optimizely says no iPhones were taken over. That is reassuring, but it is a statement about what has been observed, and the exploit kits involved are real and capable against unpatched devices.


Timeline

All times UTC.

When What happened
September 17 The same injector first appears on Turkish websites
Around November 13 A small loader is found in scripts from Kissmetrics, StorifyMe, and Minute Media, leading to the same iOS-targeting infrastructure; how it got there is unclear
November 16, 04:21 First Optimizely injection observed, in the BBC's snippet, loading from webshell[.]lol
November 16, 13:43 Card skimmer added
November 16, 17:13 to 17:57 Coruna iframe added for visitors on iOS 18.6 or earlier
November 16, 19:25 About $7,040 has reached attacker wallets (Report URI); StepSecurity estimates about $6,600
November 16, 20:40 Mass cleanup begins
November 16, 21:52 Only TalkTalk's snippet still carries the code
November 17, 01:02 Optimizely publishes its incident notice
November 17 Optimizely confirms to BleepingComputer that the flaw was in the Experimentation account signup flow

The appearance of the same infrastructure in other third-party marketing scripts before the Optimizely injections suggests this was a broader campaign against the web's marketing and analytics supply chain, with Optimizely as its largest target so far.


Indicators of compromise

The following indicators have been published by researchers (defanged):

Type Indicator Notes
Domain 1688360[.]cc Payload host; paths /track.js, /card.php, /hit.php
Domain webshell[.]lol Injector, first seen 04:21 UTC
Domain 5vobauua3ojpklx[.]icu iOS exploit stage
Domain eil786npto9u5bq[.]icu iOS exploit stage
IP 136.113.106[.]136 Attacker infrastructure
SHA-256 f7b84b9a94f749f8ced3f0112d0a66894e8bbc3b4aa5592021d1b8976a9bad40 track.js, final version
SHA-256 617df364120adc47fb4f55dcc0c69a9b6c9e108a92c60bc2b64a266254ab02f3 track.js, earliest captured

Cryptocurrency wallets

BleepingComputer extracted these attacker wallet addresses from the malicious track.js, with activity as of November 16. Some of these wallets were active before the attack, so the totals overstate what was stolen during this campaign; Report URI's figure of about $7,000 carries the same caveat.

Currency Wallet address Received Transactions
Bitcoin (BTC) 19g9rZcPeiz8rUvyWpTRyXXhPnjsWrc3kW 0.08298 BTC (about $5,380) 7
Litecoin (LTC) ltc1q3wp7g6ds0ggg6fl6kncpa4sepn0q3qfacgred5 17.02 LTC (about $762) 3
TRON (TRX / USDT) THvs6p56XkUiMkJeSc51cAP64u2RsTYqxa 662.9 USDT + 184 TRX (about $719) 50
Ethereum (ETH) 0xC4026b06D82b0F0B363eda0642D61CfbDc17884e 0.074 ETH (about $100) 6
BNB Smart Chain (USDT) 0xC4026b06D82b0F0B363eda0642D61CfbDc17884e 82 USDT (about $82) 2
Solana (SOL) FVtL8NMYS9PA1v3HE67XUGCtGA9JaZUvqav2fU2LrV5n 0 0
TON UQAgM1VqSZQiJLSG5w-ovurq5CZdxLEK3_7CMmwALTA5e09p 0 0
TON (raw format) 0:2033556a49942224b486e70fa8beeaeae4265dc4b10adffec2326c002d30397b 0 0
Bitcoin Cash (BCH) qzkaaf8n6grmpr0d9lx8xutn3r9lk722f566hhpe58 0 0
Dogecoin (DOGE) DBEGrxRF7KMSuhTS4c1MFviStU6efdMnu5 0 0

Organisations that process crypto payments can screen outgoing and incoming transactions against these addresses.

Cynet advises against blocking cdn.optimizely.com or the affected news sites outright. The host is legitimate, the malicious code has largely been removed, and a broad block will break a large number of sites without addressing the embedded-payload variant.


What you should do

1. Check your Optimizely projects now

  • Review Project JavaScript and any experiment-level custom code in every Optimizely project you own. Search for 1688360, atob(, and any script you do not recognise.
  • Fetch the snippet your site actually serves, from the CDN and from any self-hosted or proxied copy, and scan it. Record the Last-Modified header so you have a reference point if it changes again.
  • Do not assume a clean result today means you were clean on Monday. Most snippets were cleaned by Monday evening.

2. Lock down who can change your scripts

  • Audit every user, API token, and integration with write access to your Optimizely projects. Remove what is not needed and rotate credentials and tokens.
  • Ask Optimizely directly which of your projects were modified, when, and by what method. Its public notice does not say, and you will need that answer for any incident record or regulatory assessment.
  • Given the Smartsheet re-infection, keep checking your snippets over the coming days rather than treating this as a one-time review.

3. Hunt in your logs

  • Search proxy, DNS, firewall, and EDR logs from 08:00 UTC on Monday, November 16 onward for the domains and IP above. Report URI's earliest Optimizely injection was at 04:21 UTC, so extend the window if your logs allow.
  • Pay particular attention to employee devices that visited affected sites; Cynet first spotted this campaign through EDR alerts on browser connections to the attacker's server.

4. Assess customer impact

  • If your site was affected and takes card payments through its own form fields, assume card data entered on Monday may have been captured. Sites using a payment provider's hosted iframe are in a better position, but confirm how your checkout is built.
  • Rotate any secrets stored in cookies not marked HttpOnly, and invalidate sessions that may have been exposed.
  • If your site handles cryptocurrency payments or displays wallet addresses, review transactions from Monday for redirected funds, and check them against the wallet addresses listed above.
  • Consider following Salesforce's lead and disabling Optimizely scripts temporarily on sensitive properties until you have confirmed your projects are clean and Optimizely has answered your questions. Weigh this against the business impact of pausing live experiments.
  • Work with legal and privacy teams on whether customer notification is required. In many jurisdictions, skimmed card data or stolen session tokens will meet the threshold.

5. Reduce your exposure to the next one

  • Keep third-party marketing and analytics tags off checkout and login pages entirely. Use your payment provider's cross-origin hosted fields for card entry.
  • Deploy a strict Content Security Policy with tight script-src and connect-src directives. Note that allowing cdn.optimizely.com alone would not have stopped this, and strict-dynamic or unsafe-eval can let the payload run. The value of a strict connect-src here is that it can stop stolen data from leaving the page, even when the script itself comes from an allowed host.
  • Use client-side script monitoring that alerts when the content of a third-party script changes, not only when a new script appears.

6. Tell your people

Advise staff, and customers where appropriate, to:

  • Update iPhones and iPads to iOS 18.7.3 or later, or iOS 26.3 or later.
  • Check cryptocurrency payment addresses against the intended recipient, and verify them on a hardware wallet's screen rather than trusting copied text.
  • Watch card statements, contact the card issuer about anything suspicious, and consider replacing cards entered on affected sites on Monday.
  • Watch for unusual activity on any account they were logged in to while visiting an affected site.

A note on the pattern

This incident follows a familiar shape. In 2024, the polyfill.io domain was sold and began serving malicious code to more than 100,000 websites that had embedded it. The common thread is that the affected organisations did nothing wrong in their own environments. They loaded a script from a trusted provider, and that provider's integrity became their integrity.

Scott Helme of Report URI put it plainly: "trusted JavaScript doesn't always remain trustworthy." Marketing and experimentation tools are a particularly attractive target because they are designed to run arbitrary code on every page, they are often managed by marketing teams rather than security teams, and they are typically allowed by default in security policies. A single write to one field in one vendor's platform reached the visitors of some of the largest websites in the world within hours.

It is also worth noting that this is Optimizely's second security incident this year. In February, the company disclosed that a threat actor had accessed some of its systems through a voice-phishing attack, though that intrusion was limited to business contact data. The two incidents appear unrelated, but they are a reminder that vendor risk assessments should be revisited when a vendor's own security posture changes, not only at contract renewal.

For most organisations, the practical lesson is to know exactly which third-party scripts run on your pages, who can change them, and how quickly you would know if their content changed. If the answer to that last question is "when a researcher publishes a list," it is worth fixing.


Summary

Campaign OPTIMJACK
Vendor and product Optimizely Web Experimentation
CVE None assigned
Entry point Vulnerability in the Experimentation account signup flow (confirmed by Optimizely); technical details not published
Injection point Project JavaScript field, served from cdn.optimizely.com
Scale At least 188 snippets across 178 organisations, from a sample of 1,611; true number likely higher
Named victims Includes BBC, CNN, Fox News, Zoom, Salesforce, GitLab, IKEA, HSBC, Equifax, Mutual of Omaha, Air France-KLM, Smartsheet, TalkTalk, UNICEF USA, and Optimizely's own site
Exposure window November 16, 2026, 04:21 to roughly 21:52 UTC for most sites; Smartsheet re-infected after cleanup
Payloads Crypto wallet clipper; card skimmer; cookie and screenshot theft; Coruna and DarkSword iOS exploit kits
Confirmed losses About $6,600 to $7,040 received by attacker wallets, some of it predating the attack; card theft volume unknown
iOS compromise Delivery attempted; no confirmed compromise of a physical device
Vendor response Vulnerable path blocked and fixed, code removed, sessions ended, tokens revoked, access reset, code delivery blocked at Cloudflare, affected customers notified
Key IOCs 1688360[.]cc, webshell[.]lol, 5vobauua3ojpklx[.]icu, eil786npto9u5bq[.]icu, 136.113.106[.]136, plus ten attacker wallet addresses
Discovered by Report URI (Scott Helme), Cynet; analysis by StepSecurity and Hudson Rock

OPTIMJACK did not need to break into any of the companies it affected. It only needed to change one trusted script that they all chose to load. If you run Optimizely Web Experimentation, check your projects and logs today, rotate access, and get a clear answer from Optimizely about what was changed in your account.

If you need help reviewing your third-party scripts, hunting for these indicators in your environment, or assessing whether customer notification is required, get in touch.

← all insights