Optimizely Hack Poisons 178 Major Websites

An attacker exploited a vulnerability in Optimizely’s web experimentation platform to plant malicious JavaScript on its customers’ websites — hitting 178 organizations including BBC, CNN, Zoom, Salesforce, IKEA, HSBC, and Equifax. The multi-purpose malware stole cryptocurrency, skimmed payment cards, and delivered iOS exploit kits to visitors, all from Optimizely’s own trusted CDN.

What Happened

On November 16, 2026, an attacker exploited a vulnerability in Optimizely’s Web Experimentation platform — a widely used website optimization tool — and injected malicious JavaScript into the scripts that Optimizely serves to its customers’ sites. Because companies embed Optimizely through a JavaScript snippet loaded directly from Optimizely’s own content delivery network, visitors’ browsers treated the malicious code as fully trusted.

Security researcher Scott Helme of Report URI uncovered the attack and confirmed that 178 organizations were serving the poisoned scripts to their visitors. The list reads like a who’s who of the internet: Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger, and NBC’s Peacock streaming service, among many others.

Optimizely has since confirmed the incident. In a status update, the company said it identified the vulnerability, removed the injected code from affected projects, secured compromised user accounts, and put additional protections in place to prevent unauthorized code from being delivered in the future.

How the Attack Worked

The attacker planted the malicious code inside the “Project JavaScript” field within Optimizely customer projects. This is the mechanism Optimizely provides for customers to run custom experimentation scripts on their sites — and it meant the malware was delivered straight from Optimizely’s own CDN infrastructure, blending seamlessly with legitimate traffic.

Helme’s team at Report URI examined 1,611 Optimizely snippets and found 188 of them carrying the malware, linked to 178 distinct organizations. He believes the true victim count is higher, as his team could only analyze the snippets they managed to capture in time.

The injected code loaded a secondary script from a domain that was registered on the morning of the attack. Over the course of the day, the payload was rebuilt 26 times, with the attacker progressively adding features and new cryptocurrency wallet addresses — a sign of a hands-on-keyboard operation that was actively refining its toolkit in real time.

A Multi-Purpose Malware Kit

This was no one-trick payload. According to Helme’s analysis, the malware was a comprehensive attack kit that performed several functions simultaneously:

Cryptocurrency theft: The code silently swapped cryptocurrency wallet addresses displayed on affected sites with the attacker’s own addresses, redirecting any transactions to their wallets. Verified blockchain transactions confirm that crypto was successfully stolen.

Payment card skimming: A card skimmer component, which arrived in the afternoon on the first day, targeted card numbers, expiry dates, and CVV codes entered into payment forms hosted directly on affected sites. Cards entered into externally hosted payment iframes were not in scope.

Cookie and session theft: The malware exfiltrated cookies and captured screenshots from victims’ browsers, potentially giving the attacker session tokens and visual intelligence on what users were doing.

iOS exploit delivery: On iPhones and iPads running iOS 18.6 or older, the malware delivered two known iOS exploit kits — Coruna and DarkSword — which are capable of stealing keychain data, seed phrases, and private keys. Affected pages on these devices also reloaded every 60 seconds, a behavior some visitors may have noticed.

Banks and financial institutions made up the largest group of victims, with 32 organizations hit, followed by technology companies and retailers — sectors where payment data and high-value credentials are most likely to flow through web sessions.

Not an Isolated Incident

The Optimizely compromise was part of a broader campaign. Researchers at Hudson Rock and an independent investigator identified poisoned scripts being served through three additional web services: Kissmetrics, StorifyMe, and minute-ly. The earliest tampered responses from these services were recorded as far back as November 13, suggesting the campaign was already underway days before the Optimizely injection was discovered.

The different services share hosting infrastructure and loader code, though researchers caution that this alone does not prove a single operator was behind all of them or that the same entry point was used in every case.

One of the attacker’s cryptocurrency wallets had been active since March 2026, and the same injection code first appeared on Turkish websites in September. Despite the scale of the operation, the financial haul from the crypto-swapping component was modest — around $7,040 — before the funds were moved to exchanges.

Most Victims Lacked Adequate Defenses

One of the most striking findings from Helme’s research was how few organizations had defenses in place that could have mitigated the attack. His team crawled the affected organizations’ websites to evaluate their Content Security Policy (CSP) configurations.

Only 24 of the 178 affected organizations — roughly 15% — had a CSP strict enough to have blocked the malicious payload. Another 68 organizations had no CSP at all. In 50 cases, the existing policy actually permitted the malware to execute and send stolen data back to the attacker.

Making matters worse, even organizations that had whitelisted Optimizely’s domain in their CSP got no protection, because the malware was being served from that exact domain. Optimizely’s own documentation recommends a configuration that enables JavaScript’s eval function — a setting that allowed a backup copy of the malicious code to execute on HSBC’s site.

Takeaways and What to Watch

This attack is a textbook example of a supply-chain compromise through trusted third-party JavaScript — the kind of risk that is growing across the web as organizations embed dozens of external scripts on their pages. Here is what defenders should focus on:

Audit your third-party scripts. Catalog every external JavaScript dependency your site loads and assess what level of access each one has. If a provider can inject arbitrary code into your pages, that is a critical trust boundary.

Implement Subresource Integrity (SRI). Where possible, pin external scripts to known hashes so that any modification is blocked by the browser before it executes.

Tighten your Content Security Policy. Do not rely on domain-based allowlists alone. Use nonce-based or hash-based CSP directives, and avoid configurations that permit eval or unsafe-inline.

Monitor for behavioral anomalies. Real-time JavaScript monitoring tools can detect when a trusted script suddenly starts making requests to unfamiliar domains or accessing form fields it normally would not touch.

Patch iOS devices. The exploit kits delivered in this campaign targeted devices on iOS 18.6 and earlier. Users and organizations should ensure all Apple devices are running iOS 18.7 or later.

As Helme concluded in his analysis, trusted JavaScript does not always remain trustworthy. A website can be fully compromised through a third-party dependency without its own code or infrastructure ever being breached. In 2026, managing the security of what runs in your visitors’ browsers is just as important as securing your own servers.

Sources: Help Net Security, Report URI (Scott Helme), Hudson Rock, Optimizely Status, Reuters

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.