Optimizely has confirmed that an attacker exploited a vulnerability in its Experimentation platform to insert malicious JavaScript into customer projects. The immediate response for website owners is to verify the code their visitors actually received, then investigate payment and session exposure separately from the vendor's cleanup.
Scott Helme's Report URI investigation, published November 16 and updated November 17, found 188 infected JavaScript snippets belonging to 178 organizations in a sample of 1,611 snippets. Those are observed research counts, not a complete customer-impact assessment or a count of compromised visitors.
What is confirmed, and what has changed
In a status notice posted November 17 at 01:02 UTC, Optimizely said an unauthorized party exploited an Experimentation vulnerability and added malicious JavaScript to some customer projects. The company said it fixed the vulnerability, removed the injected code, secured affected accounts and added protections against unauthorized code delivery. The notice remained in monitoring when reviewed.
Report URI traced the malicious additions to Project JavaScript, custom code delivered inside a customer's Optimizely snippet. The trusted delivery hostname and certificate did not make that added code safe. The researcher explicitly distinguished the incident from CDN cache poisoning: the delivery service was serving modified project content.
The researcher's November 17 update says all infected Optimizely files it observed were clean by November 16 at 23:11 UTC and had remained clean. That observation supports containment of the sampled delivery path. It does not establish that every affected browser, session or transaction has been recovered.
Why browser exposure matters
The analyzed payload could replace cryptocurrency wallet addresses, including through clipboard interactions; capture card information entered in ordinary page fields; and collect accessible cookies and screenshots in documented circumstances. Report URI says the skimmer could not reach card fields inside hosted payment iframes. That limitation should inform exposure review, not become a blanket assurance that an entire checkout or browser session was safe.
The report also documents delivery of Coruna and DarkSword exploit-kit code to iPhones and iPads. Delivery is not proof that every visiting device was successfully exploited. The researcher could not determine how many cards, cookies or screenshots were stolen, or how many iOS devices were infected. Wallet transaction receipts likewise are not a reliable total for losses caused by this incident.
Actions for website owners and their providers
- Find the actual dependency. Inventory Optimizely Web snippets across production, preview, login and payment-related pages, including copies served through your own domain. Ask the web agency or managed provider for project identifiers and deployment locations, not just confirmation that the main homepage looks normal.
- Preserve and compare. Retain relevant snippet versions, response headers, timestamps, deployment history and available browser-security reports. Review Project JavaScript and experiment custom code against authorized changes. A search for one domain is insufficient: the researcher found encoded loaders and an inline fallback.
- Verify clean delivery. Inspect the script and its dependent requests as delivered to a fresh browser session. If integrity cannot be established, consider temporarily disabling the integration while evaluating business impact. Vendor remediation and a clean dashboard are useful evidence, but neither replaces checking the page visitors load.
- Scope data and session recovery. Determine which pages and users were exposed during the observed November 16 activity, and whether sensitive data was present in page fields or JavaScript-accessible cookies. Invalidate potentially exposed sessions and rotate affected secrets according to scope. Review project users, API tokens and integrations for unauthorized access.
- Improve browser controls deliberately. Test restrictive script, connection and frame policies, with reporting, before enforcement. Review dependence on unsafe evaluation and delegated script trust. Blocking a collection endpoint may limit theft without stopping in-page address replacement; a CDN allowlist alone cannot establish script integrity.
Actions for visitors and lean security teams
If staff visited an affected site on November 16, review relevant cryptocurrency destinations against a separately verified intended address. Monitor card activity and contact the issuer about suspicious transactions. Ensure iPhones and iPads run current supported security updates. Treat a visit as an investigation lead, not automatic proof of device compromise; escalate suspicious device or account activity through the established incident-response process.
For government contractors, involve the security and contractual-response owners if the review identifies possible exposure of sensitive work information. This reporting alone does not establish a reportable contractual incident or compromise of a customer's internal network.
Bulwark assessment
The central failure was delegated execution trust: code authorized through a shared service ran inside customer webpages. Recovery therefore has two separate acceptance tests: the supplied script is clean, and the access or information exposed before cleanup has been addressed. Closing only the first leaves the business risk unresolved.
Primary source: Optimizely Web Experimentation security-incident notice.
Original research: Scott Helme, Report URI: Optimizely malware investigation and November 17 update. Consult the original indicator table for roles and context; customer sites and the legitimate delivery service are not malicious indicators.
Secondary context: BleepingComputer's corroborating reporting and vendor response.

