BlackTree Security · Infrastructure · Automation · AI

Optimizely Scripts Carried Malware Into Major Websites

The websites had not necessarily changed. The script they trusted had.

On 16 November, malicious JavaScript reached visitors through Optimizely Web, an experimentation service used to test and personalise web experiences. The code arrived inside customer snippets delivered from infrastructure those sites already expected to load. That made the incident more consequential than a suspicious script added to one compromised website: a shared publishing path carried hostile behaviour into many unrelated brands.

Report URI’s investigation found 188 infected snippets associated with 178 organisations. HSBC web properties appeared in its research. Those findings describe infected website snippets observed by the researchers. They do not establish that 178 corporate networks were breached, that every visitor was harmed, or that every named organisation lost data.

Optimizely confirms a vulnerability was exploited

In an update dated 17 November, Optimizely said an unauthorised party exploited a vulnerability in Experimentation and used it to add malicious JavaScript to some customer projects. The company said it had fixed the vulnerability, removed the injected code, taken steps to secure affected user accounts and added protections intended to prevent unauthorised code delivery.

The notice did not give a customer count, identify the vulnerability in technical detail or say when the attacker first obtained access. It also did not quantify visitor exposure or confirmed data loss. In its 17 November notice, the incident was marked as monitoring rather than resolved.

That vendor account answers one central question left open in the earliest research: the activity involved exploitation of Optimizely Experimentation. It does not answer every question about the initial access, the exact projects reached or the complete period during which a visitor could have received hostile code.

The malicious code travelled inside a legitimate snippet

Optimizely Web customers load a JavaScript snippet that contains the service’s code, experiments and an optional Project JavaScript field. Report URI found that the attacker had written a loader into that customer-controlled field. The rest of the snippet remained in place, and the resulting file could still be served from an Optimizely domain with a valid certificate.

This distinction matters. Report URI did not describe a cache-poisoning attack against the content delivery network. Nor does finding the injected script on a company’s web property prove that the company’s origin server was compromised. The abused trust sat in the route through which authorised experimentation code could be changed and distributed.

The observed payload could interfere with cryptocurrency payments, collect card details from page forms and steal other browser-accessible information. Its capabilities do not establish the number of visitors harmed. Optimizely had not published a confirmed total for visitor exposure or financial loss.

Exposure needs a dated reconstruction

A clean script downloaded today cannot establish what a visitor received yesterday. Investigation should preserve the period being assessed and distinguish evidence retained during the incident from a later clean copy. A business also needs to identify which of its customer journeys used the affected project, rather than treating every page under its domain as equally exposed.

Payment, identity and privacy teams may need different records from the same browser session. A published script history can identify a potential exposure window; deciding which people require support or notification needs further evidence about the information available on those pages and the behaviour actually observed. Teams should document those decisions and revisit them when the investigation changes.

A trusted host is not a permanent security decision

The immediate defensive question is not simply whether an organisation uses Optimizely. It is which pages load the snippet, who and what can alter project code, which tokens and integrations can publish changes, and whether defenders can reconstruct what visitors received during the affected period.

Content Security Policy can reduce impact, but allowlisting a trusted provider’s host is not sufficient when that provider delivers the altered code. The research showed that restrictive script and connection policies could limit the payload, while incomplete policies left routes open. Script-change monitoring, controlled publishing rights and outbound destination restrictions address different parts of the problem.

Organisations investigating their own exposure need evidence from the browser layer: retained snippet versions, project change history, account and token activity, Content Security Policy reports, web telemetry and the location of sensitive forms. The incident shows why a third-party script inventory must describe more than a domain name. It must identify who can change the code, where it executes and how quickly the organisation can stop trusting it.

Leave a Reply

Your email address will not be published. Required fields are marked *