Four engineering questions for investigating third-party software exposure.
On 16 November 2026, attackers exploited a vulnerability in Optimizely Web Experimentation, allowing them to impersonate customer customers and inject malicious JavaScript into legitimate website scripts.
In a recent development, researchers at Report URI identified 188 affected scripts across 178 organisations, including major international brands. The malicious code was capable of stealing payment information, manipulating cryptocurrency transactions and targeting browser-accessible information.
Optimizely has since fixed the vulnerability and removed the injected code. Its forensic investigation remains ongoing.
But containing the incident doesn't automatically establish what happened on every affected website.
If your application loads a compromised third-party component, how do you establish what it could access and whether anything was actually exposed?
Here are four questions engineering teams should investigate.
Start by identifying which applications and pages loaded the affected component.
For the Optimizely incident, the reported exposure window was 07:51β22:24 UTC on 16 November, although individual customer exposure windows may differ.
It has been revealed that optimizely has confirmed that Web Experimentation was affected, rather than its other experimentation products. Customers should confirm their individual exposure status and window directly with Optimizely.
What to establish: Which pages loaded the script, when it was active and whether visitors could have received the malicious version.
Historical deployments, tag-manager configurations, archived script responses and browser monitoring records can help reconstruct that picture.
Knowing that a script was present doesn't tell you what it could do.
Third-party JavaScript running in a website's context may be able to interact with page content, form fields, browser-readable storage and outbound requests, subject to browser security controls.
For this incident, researchers observed malicious capabilities including payment-card skimming and cryptocurrency address manipulation.
That makes the execution context particularly vital.
Payment details entered inside properly isolated third-party payment iframes are generally inaccessible to scripts running on the parent page. Optimizely's initial investigation also identified that distinction.
What to establish: The information and functionality available to the script, not simply whether the dependency existed.
Importantly, browser-level exposure does not automatically mean an organisation's backend infrastructure was compromised.
This is where understanding the wider architecture becomes useful.
A third-party script may be supplied by an external organisation but execute within the context of your own website. That creates a trust relationship worth examining.
Review the controls that determine what it can reach and where information can travel.
According to recent reports, content Security Policy (CSP): Does your policy restrict executable scripts and outbound connections? Are directives overly permissive?
Data flows: Which payment, authentication and customer-information workflows operate on affected pages?
Isolation: Are sensitive interactions separated from third-party scripts through appropriate browser and application boundaries?
A dependency might have legitimate access to an application while still creating exposure if that access is broader than necessary.
What to establish: Which safeguards were present, whether they were effective and what an attacker could potentially reach through the affected component.
Industry observers note that an architecture diagram can help identify potential exposure. It cannot prove that malicious code executed or that information was stolen.
Incident investigations need historical evidence.
It's useful to separate findings into three categories:.
The absence of suspicious logs is not necessarily proof that nothing happened. Some browser-side activity may not appear in server-side monitoring.
For affected Optimizely customers, preserving relevant logs, reviewing payment activity and following the vendor's incident guidance should be immediate priorities.
Industry observers note that traditional incident response remains essential. Teams still need to contain threats, preserve evidence, investigate indicators of compromise and assess any notification obligations.
Architectural understanding supports that work by showing how an affected component relates to the wider application.
It helps answer questions such as:.
At Veriom, we're building technology that maintains a living model of software architecture across code, cloud and infrastructure, helping engineering teams understand dependencies, data flows and trust boundaries.
According to recent reports, that context can help teams investigate and prioritise structural risks. It doesn't replace browser forensics, incident response or specialist security testing.
When a trusted dependency is compromised, start with four questions:.
According to recent reports, A dependency inventory tells you what software you're using. Understanding how that software interacts with the rest of your system helps you assess what a compromise could actually affect.
For teams looking to improve their architectural visibility, explore Veriom.
