A supply-chain compromise does not have to touch your own web property to expose your users. On November 16, 2026, malicious JavaScript was injected into Optimizely Web Project JavaScript and delivered through trusted Optimizely-backed snippets on major third-party websites.
What makes this attack interesting is that an impacted organization doesn’t have to use Optimizely itself. A user only has to browse to an affected site.
This gave us a concrete hunt objective. We needed to identify any users who touched the campaign, determine how far the activity progressed, and figure out whether anything required follow-up. Nebulock let us turn newly published intelligence from Scott Helme and Report URI into an environment-specific hunt in minutes.
How the attack works Optimizely Web lets websites load project-specific JavaScript for experimentation and A/B testing. In this attack, an unauthorized party gained the ability to add malicious code to the Project JavaScript of multiple customer projects. The trusted Optimizely-backed snippet still looked normal to the browser, but the injected loader then reached out for the campaign’s track.js payload.
Delivery flow: User visits affected third-party site → trusted Optimizely-backed snippet loads → malicious loader in Project JavaScript runs → 1688360[.]cc/track.js loads → card skimming / crypto-address swapping / cookie + screenshot collection → iOS devices may branch to exploit-kit infrastructureThe important part for defenders is that their organization does not need to own the affected website. An employee can be exposed simply by visiting a third-party site that loaded the compromised Optimizely code.
For the underlying incident mechanics, timeline, and published IOCs, see Scott Helme and Report URI’s technical analysis and Optimizely’s incident update .
From fresh intel to a hunt in minutes with Nebulock If a SOC was hunting this threat manually, they would start by pulling the IOCs from the reporting, building queries, and running them across the SIEM, EDR, DNS, and other available network telemetry. AI can help write the queries, but there is still plenty of copying, pasting, tool hopping, and rerunning searches.
Then each new hit puts another set of pivots on your plate. An exact campaign-domain hit gives evidence of exposure, but you still want to know how far the activity progressed and tie it back to the actual endpoint and user. Depending on the telemetry source, that may mean correlating an IP, hostname, user, browser activity, and any follow-on campaign traffic.
The Nebulock ACE platform dramatically reduces the amount of steps that go into a threat hunt. Analysts can start a hunt a few different ways, like with an advisory or threat-intelligence link, a hunter-written hypothesis, a suggested hunt, or some combination of those inputs. For this campaign, the goal is the same regardless of how the hunt is initiated: scope the campaign infrastructure, execute the hunt across available telemetry, and correlate any matches back to affected entities.
A natural-language starting point can be as simple as:
Using this report [report URL], hunt for endpoint exposure to this campaign and give me the affected users or hosts that may need follow-up. From there, Vespyr autonomously hunts the scoped campaign infrastructure, identifies the endpoints and users associated with any matching activity, and characterizes the surrounding evidence rather than stopping at a raw IOC match.
An IOC hit is a good thread to pull on, but it is not the whole story. Vespyr pulls those pivots together: what matched, how far the activity progressed, which endpoint and user it belongs to, and whether there is actually something to act on. No hit hunts are useful too. If the hunt checks the available telemetry and finds no evidence of exposure, that still closes an important question.
The resulting hunt report gives analysts the executive summary, findings and supporting evidence, MITRE ATT&CK context, investigated behaviors, recommendations, and the underlying LOCK/ABLE hunt structure all in one place.
Top of a redacted threat hunt report Agentic and suggested hunting in Nebulock This was the kind of threat intel our DE/TH team did not need to overthink. It was new, there were plenty of concrete observables to work from, and the exposure model meant it was worth checking even in environments that do not use Optimizely directly. So we created a threat hunt that was run against all Nebulock customer environments.
This hunt has been added as a one-click hunt within the platform and as an autonomous Vespyr hunt, so teams can run the same scoped hypothesis without rebuilding the hunt from scratch.
Suggested hunts in Nebulock Title: Optimizely Supply-Chain Skimmer (Magecart) - Endpoint ExposureDescription: A client-side supply-chain compromise disclosed 2026-11-16 injected malicious loader code into Optimizely Web Project JavaScript. That loader delivered a Magecart-style track.js payload capable of payment-card skimming, crypto-address swapping, cookie and screenshot collection, and routing iOS devices toward exploit-kit infrastructure. This hunt identifies endpoints and users that contacted campaign infrastructure, whether activity progressed beyond one-off exposure, and whether collection/exfiltration paths or iOS routing were observable.Hunt Directive (agent instructions, carries the IOCs): Hunt for endpoint exposure to the Optimizely / Magecart client-side supply-chain campaign disclosed 2026-11-16. Search DNS, HTTP, and network-connection telemetry over the last 14 days for contact with the campaign infrastructure and report affected endpoints and users with first and last seen. Primary skimmer and crypto-clipper C2: 1688360[.]cc (paths /track.js, /card.php, /hit.php, /html2canvas.min.js). Injector infrastructure: webshell[.]lol (/get.js) and linkforge[.]cc. iOS exploit-kit routers (.icu): 5vobauua3ojpklx[.]icu, eil786npto9u5bq[.]icu, l7bcsnwc7nzr3uo[.]icu, je4ft3gn8vxudum[.]icu, q1ls1ho9x7wsxlu[.]icu, ios-hack[.]icu, p55abnw2pnsz0bo[.]icu, 1y69gy2mdzaltz1[.]icu, gm38duxoihbmy5o[.]icu. Associated IPs may corroborate but must NOT be alerted on alone because they are shared cloud ranges: 136[.]113[.]106[.]136, 34[.]96[.]213[.]201, 34[.]87[.]36[.]58. track.js SHA-256: 814a963c0da480e2449d9b6664480156e1cd6d731014117401bf635badec3a0f and f7b84b9a94f749f8ced3f0112d0a66894e8bbc3b4aa5592021d1b8976a9bad40. For each endpoint that resolved or connected to this infrastructure, determine whether the skimmer actually beaconed (repeated contact, or the /card.php and /hit.php exfil paths), which user and browser were involved, and whether any iOS device was routed to an exploit kit. In-page markers: window._k7, window.__k7b, data-channel-code=02111110, telemetry channel code BIGNIAO. The malicious code was injected into Optimizely Web Project JavaScript served from the trusted cdn[.]optimizely[.]com, so treat contact as browser-delivered from a third-party website, not necessarily a compromise of this organization's own web property.The shared-cloud IP distinction is important. An IP match by itself is not enough context to call campaign exposure: cloud infrastructure can be shared or reassigned, so the domain, timing, browser activity, and other surrounding evidence are what make the hit meaningful.
Next steps following a hunt With the hunt results, security teams can quickly:
Preserve the contact time, user, browser, and surrounding browsing activity Confirm what site the user had open and whether they entered payment-card data, credentials, or cryptocurrency information Rotate or monitor anything that may have been exposed based on the user activity and available evidence Block and monitor the campaign domains Check the published [.]icu router infrastructure separately for iOS exposure where mobile telemetry exists Whether the hunt returns matches or nothing at all, the context should not disappear when the hunt closes. Nebulock’s TRACE Graph keeps the hunt, evidence, and environmental context available to future hunts, detections, investigations, and triage so the next hunter or agent does not have to rebuild the same picture from scratch.
If your organization itself uses Optimizely Web, that is a separate review of its own Project JavaScript and affected projects. This endpoint hunt is answering a different question: whether your users were exposed while visiting someone else’s affected site.
Get a live demo of Nebulock to see how quickly you can go from threat intel to understanding impact.