What happened
On 16 November 2026, Report URI observed malicious JavaScript delivered through Optimizely Web Experimentation to 178 organisations after finding 188 infected snippets among 1,611 examined. The injected Project JavaScript loaded track.js from 1688360.cc while the parent snippets continued to arrive from cdn.optimizely.com. The payload could swap cryptocurrency addresses, skim card details from first-party forms, collect non-HttpOnly cookies and screenshots, and route iOS devices to the Coruna and DarkSword exploit kits. The attacker therefore gained browser-level access on affected pages without needing to compromise each organisation’s origin infrastructure.
The published payload indicators include 1688360.cc, 5vobauua3ojpklx.icu, eil786npto9u5bq.icu, SHA-256 814a963c0da480e2449d9b6664480156e1cd6d731014117401bf635badec3a0f and SHA-256 f7b84b9a94f749f8ced3f0112d0a66894e8bbc3b4aa5592021d1b8976a9bad40. Report URI says the exploit-kit comments treated iOS 18.7.3 and later and iOS 26.3 and later as patched. Its analysis also found that hosted payment frames were outside the card-skimming code’s reach, while card fields implemented directly by affected sites were readable.
Report URI observed all infected Optimizely files clean by 23:11 UTC on 16 November 2026 and recorded the main domain 1688360.cc on clientHold at 00:37 UTC on 17 November 2026. Optimizely posted at 01:02 UTC on 17 November 2026 that an unauthorised party had exploited a vulnerability in Experimentation, added malicious JavaScript to some customer projects, and that the company had fixed the vulnerability and removed the injected code. Optimizely did not publish an affected-project list, root-cause report or visitor count in the cited status update. Attribution posture: Neither Optimizely nor Report URI identified a responsible actor in the cited sources.
Why this matters now
This was not simply a compromise of an organisation’s own application code. The browser received attacker-controlled behaviour from a previously approved dependency, over an expected vendor domain and certificate. Traditional server patching, endpoint controls and origin integrity checks therefore cannot establish whether customer sessions remained safe. Security leadership needs client-side evidence showing what executed, where it executed and what destinations the browser reached.
Vendor cleanup removes the active delivery path but does not determine whether card details, session material, cookies, screenshots or cryptocurrency transactions were exposed before removal. Closure depends on each customer’s page inventory, script history, browser telemetry, CSP reporting, proxy and DNS evidence, form architecture and session design. Organisations lacking those records have an assurance gap, not evidence of absence.
The incident crosses security, fraud, privacy, digital commerce and customer communications. Organisations using first-party payment forms face a different exposure from those using isolated hosted frames. Authentication pages carrying non-HttpOnly cookies create a separate identity question. Older iOS visitors introduce endpoint compromise risk. One vendor incident therefore requires several scoped decisions rather than a single platform-wide declaration.
The decision for security leaders
Make digital-platform owners prove exposure at project and page level. A vendor statement that injected code was removed does not establish which revision each customer served, whether an affected page contained sensitive forms or cookies, or whether the browser reached attacker infrastructure. Require a signed exposure inventory and preserved evidence before closing the incident.
Separate script remediation from compromise and data-impact decisions. Security operations should determine execution and egress; identity teams should determine session exposure; payment and fraud teams should assess first-party card forms and transaction intersections; privacy and legal should decide notification. Combining these into a single patched-or-not status obscures materially different risks.
Change the standing control model for third-party browser code. Require named ownership, an approved page inventory, change detection, restrictive script and connection policies, browser-side telemetry, emergency disablement and contractual incident evidence. The control objective is not merely trusting a supplier; it is detecting when approved code begins performing an unapproved action.
Evidence of closure
- Signed inventory of Optimizely project IDs, pages, owners and approved script hashes.
- Preserved evidence showing which script revision each critical page served during the active window.
- Validated telemetry showing no connections to published attacker domains, or a documented incident case.
The Security.io assessment
The delivery and remediation are confirmed through Optimizely’s status page and direct technical observation. The affected scale is a measured research sample, not a complete customer census. Report URI could only assess snippets it found, while Optimizely described some customer projects without publishing the boundary. The evidence supports urgent customer-specific scoping but not a conclusion that every Optimizely customer or every visitor was affected.
Cleanup materially reduces ongoing delivery risk, but it does not answer historical exposure. Domains becoming unreachable also cannot retroactively demonstrate that earlier requests, card submissions, cookie collection or exploit attempts failed. Organisations that lack retained CSP reports, browser monitoring, script versions or detailed proxy logs may need to record an assurance limitation rather than assert that no impact occurred.
The most consequential control failure is the assumption that a trusted script remains trustworthy after approval. This incident should move browser-executed dependencies into the same governance class as privileged software integrations: inventoried, monitored for behavioural change, constrained by policy and covered by evidence-producing incident clauses. That decision has value beyond Optimizely and should survive closure of this incident.
Questions for the morning meeting
- Can we prove which Optimizely code each critical customer journey served during the active window?
- Which payment, authentication or account-recovery pages trusted this third-party script?
- Do our contracts require incident scope, root-cause and customer-notification evidence?
- Can CSP and browser telemetry block or expose the same dependency failure?