Skip to content
Back to blog

Optimizely Script Compromise: Third-Party Checkout Skimming Explained

6 min read
Optimizely Script Compromise: Third-Party Checkout Skimming Explained

Sources checked: . FailSafe Research.

A reported compromise of a script delivered through Optimizely's web platform placed payment-skimming code in customer checkout pages. The page owners did not deploy the malicious code. A script they trusted changed underneath them. That is the defining property of a client-side supply-chain failure: the vulnerable component is not in your repository, and your normal release review never sees it.

Report URI's analysis, as relayed to FailSafe, describes approximately 178 affected organizations, with banking and finance the hardest-hit sector at about 32 organizations. Reporting also describes device-conditional behavior, including logic aimed at iOS visitors, and attempts to avoid analysis. We could not retrieve the primary Report URI pages during this review. Treat the count, sector split and named organizations as that publisher's assessment, not as figures FailSafe independently confirmed. This article does not repeat the victim list as fact.

What the reported pattern means

A third-party tag is loaded from a vendor domain and executed in the visitor's browser. On a checkout page, that code runs in the same context as the payment form. If the vendor's delivery is altered, the page can collect what the visitor enters or change how the form behaves. The site's own code, dependencies and build pipeline can all look clean.

Several properties make this class difficult to catch:

  • The script is served from a domain the site already trusts and may allowlist in policy.
  • Its content can change without a deploy, a pull request or a new release note.
  • Dynamic tag injection often bypasses Subresource Integrity checks that would apply to a fixed script tag.
  • Device-specific logic can make behavior differ between a security researcher's browser and a real customer's phone.
  • Payment data theft can occur without a visible change to the page a user sees.

Questions to ask about every checkout script

Trust assumptions and the evidence that should replace them
AssumptionWhat to verifyEvidence to keep
The vendor is trustedWhich domain serves the script, who can change it and what the contract says about incident notice.Vendor owner, script URL, review date and notification terms.
The script only does analyticsWhat data it can read from the checkout context and where it can send it.Purpose, data fields, permitted destinations and egress control.
CSP covers itWhether the policy allowlists a broad vendor host or a specific path, and whether reports are collected and reviewed.CSP rule, report endpoint, a denied-script test and report triage.
Integrity is enforcedWhether a fixed version with SRI is possible, or whether the tag is injected dynamically.Pinned version, integrity attribute or a documented exception.
Changes are reviewedWho monitors vendor script changes and how quickly a change reaches the site owner.Hash or content monitoring, alert owner and rollback path.
Payment pages are isolatedHow many third-party scripts run on the checkout page and whether a hosted payment field reduces exposure.Script inventory for checkout, isolated payment component and test result.

Controls that reduce exposure

No control makes a third-party script trustworthy. The practical goal is to reduce what a compromised script can reach and how long it can operate.

  1. Inventory by page context. Record every script on the checkout path, its owner, its purpose, the data it can read and the domains it can contact. A site-wide tag list is not enough; a checkout-page list is the relevant boundary.
  2. Constrain execution and egress. Use a Content Security Policy that allows the specific script paths you need, not whole vendor domains. Collect violation reports and review them. Confirm that an unapproved script is actually blocked, not only reported.
  3. Pin and verify where the vendor supports it. A fixed version with Subresource Integrity is stronger than a dynamically injected tag. Where dynamic injection is required, document the exception and compensate with monitoring.
  4. Detect change. Monitor the content hash of critical vendor scripts and watch vendor advisories. A change in a checkout script should create an alert with a named owner.
  5. Minimize checkout scripts. Remove tags that do not serve the payment task. Prefer hosted payment fields or an isolated iframe so the page's other scripts cannot read the card fields.
  6. Prepare the response. Know how to disable a vendor script immediately, who can do it, how to preserve browser and server logs, and how to notify customers and regulators if payment data may have been exposed.

Tests that produce evidence

These tests belong in an authorized assessment of pages you own or are permitted to test. They do not require changing the vendor's service.

Test area, safe method and the evidence to record
AreaSafe methodEvidence
Script inventoryLoad the checkout page with browser instrumentation and record every script URL and initiating owner.URL list, load order, owner and page context.
CSP enforcementAttempt to load an unapproved script in a controlled test page and inspect the block and the report.Blocked request, console or report record and policy line.
IntegrityCompare the served script with the pinned version and confirm whether an integrity check exists.Hash comparison and the tag's integrity or pinning state.
Data reachUse synthetic form values and a canary destination to see which fields a permitted script can read.Fields observed, destination attempted and whether egress was denied.
Change detectionSimulate a content change in a staging copy and confirm the alert and owner routing.Alert timestamp, recipient and rollback step.
ResponseDisable a noncritical vendor script in staging and verify the checkout still completes.Disabled state, completed test transaction and restored state.

How FailSafe can help

FailSafe tests deployed applications and the third-party scripts that run in the browser. A scoped engagement can inventory checkout scripts, verify CSP enforcement, test whether synthetic payment fields are reachable by a permitted script, and confirm that a denied destination is actually blocked. The Polymarket JavaScript supply-chain analysis covers a related pattern in a different context. The application penetration testing scope describes how a browser and API test is agreed.

We do not test a third-party platform without that provider's authorization, and we do not publish payloads or reproduction steps. If you suspect an active compromise, contact the vendor and your incident-response provider first, preserve evidence, and disable the affected script through the vendor's documented process.

What this does not establish

This article does not confirm the reported organization count, the sector split, the named victims, the data impact or the exact mechanism used to alter the vendor script. Those are Report URI's assessment as relayed to us, and FailSafe did not retrieve the primary pages during this review. Describing device-specific logic as an "iOS exploit" may overstate what was observed; the reported behavior is consistent with conditional targeting or evasion, which is not the same as a device vulnerability. Confirm current details with Optimizely, Report URI and the affected organizations before relying on them.

Questions & answers

Questions about this topic

Direct answers to common questions about this page.

No. The figure is attributed to Report URI. FailSafe could not retrieve the primary reporting during this review, so the count, sector split and named organizations are not independently confirmed here.

It reduces exposure and gives you reports, but a broad allowlist for a vendor domain can still permit a compromised script. Confirm that unapproved scripts are blocked, not only reported.

SRI helps when the vendor serves a fixed version you can pin. It does not cover dynamically injected tags or a vendor that changes content without changing the URL.

Inventory the scripts on the checkout path, identify the owner and purpose of each, and confirm how quickly each can be disabled. Then test CSP enforcement and review vendor incident terms.

Put this guidance to work

Bring the checkout pages, script inventory and policy you want tested. FailSafe can verify enforcement and produce bounded evidence.