Optimizely Signup-Flow Vulnerability Injected Malicious JavaScript Into Sites of 178 Organizations
Key Facts
- Optimizely confirmed the attacker exploited a vulnerability in its Experimentation account signup flow.
- Scott Helme of Report URI identified 188 infected JavaScript snippets associated with 178 organizations, including the BBC, CNN, Zoom, GitLab, IKEA, and HSBC.
- The payload delivers cryptocurrency clippers, card skimmers, and iOS malware.
- One snippet loads boot.js from 5vobauua3ojpklx[.]icu and sets data-channel-code="02111110" and data-cfasync="false"; the sample's decoded modules are 8,501 bytes and 312 bytes.
- Optimizely says it patched the vulnerability, removed the injected code, revoked tokens, reset access for affected accounts, and notified the organizations involved; no CVE has been assigned.
What Happened
Optimizely confirmed that the attacker exploited a vulnerability in its Experimentation account signup flow. On November 16, 2026, that flaw was used to distribute malicious JavaScript through websites operated by some of the world’s largest brands.
The path the code took is central to the story. The attacker inserted code into a field in the signup flow, and that code was then served through Optimizely’s own trusted infrastructure. Publishers that embed a third-party experimentation platform hand part of their JavaScript delivery to that platform; here, the delivery path itself carried the malicious script.
Optimizely’s Web Experimentation platform is the product named in the company’s confirmation, and the signup flow is the specific entry point it identified. The service’s role in delivering experiment code to customer sites is what connects a signup-flow flaw to pages operated by the named organizations.
Security researcher Scott Helme of Report URI identified 188 infected JavaScript snippets associated with 178 organizations. Among the named organizations are the BBC, CNN, Zoom, GitLab, IKEA, and HSBC. The payload delivers cryptocurrency clippers, card skimmers, and iOS malware, according to research published by StepSecurity.
The counts describe snippets and organizations, not visitors or sessions: 188 infected snippets across 178 organizations, with the named brands spanning media, conferencing, developer tooling, retail and banking.
Inside the Injected Script
One snippet creates a script element that loads boot.js from 5vobauua3ojpklx[.]icu and sets data-channel-code=“02111110” and data-cfasync=“false”. Both attributes sit on the same script element as the encoded bootstrap, and the domain and channel code are the artifacts a defender can search for in page source, archived copies and proxy logs.
At runtime, a wrapper reverses the encoded string, decodes it with atob(), and executes the result via new Function():
(new Function( atob(encoded.split("").reverse().join("")) ))();
The wrapper is short by design. Everything the snippet does downstream is contained in the reversed, base64-encoded string, so the visible part of the snippet carries no readable logic beyond the loader and the two attributes.
The sample’s decoded modules are 8,501 bytes and 312 bytes respectively. Because the string is reversed and base64-encoded before execution, the payload does not appear in readable form inside the snippet; the two modules exist as encoded data until the wrapper runs.
The Clipboard Component
One of the two modules handles the clipboard. It registers a capture-phase copy listener that reads W.getSelection(), applies a transform, and overwrites clipboardData with setData(‘text/plain’, o) before calling preventDefault().
The listener therefore replaces what the user copies with transformed content while the copy is in progress, working from the current selection rather than from stored clipboard data.
The analysis records two limits on that behaviour. The component does not read the clipboard’s prior contents, and the wrapped Clipboard API methods remain subject to browser restrictions. The second module, by contrast, is restricted by the outer configuration to devices classified as iOS.
Optimizely’s Response
Optimizely says it patched the vulnerability, removed the injected code, revoked tokens, reset access for affected accounts, and notified the organizations involved. No CVE has been assigned.
The sequence is worth reading literally: patch first, then removal of injected code, then token revocation and access resets, then notification of the organizations involved.
Several questions remain open. The record available for this analysis does not state how long the injected snippets were served before removal, whether each of the 178 organizations received the same modules, or how many visitors to the affected sites ran the code. Those gaps matter for scoping exposure, because the affected population is currently defined by a customer list and a snippet count rather than by a measured number of executions.
What to Watch
For defenders, the practical artifacts are the injected snippet pattern, the boot.js path on 5vobauua3ojpklx[.]icu, and the data-channel-code value 02111110. Searching for those strings in archived page source, content security policy reports and proxy logs is the direct way to test whether a given site was touched.
The broader lesson is about the boundary of hosted scripts. An organization can harden its own signup flow and still serve a third party’s experiment code to every visitor; the trust placed in that third party is what the attacker used, and it is why the blast radius is measured in organizations rather than in a single site.
The primary research is published by StepSecurity, with additional coverage referenced by Socket.
Sources & References
- OPTIMJACK: Optimizely-Hosted Scripts Loaded a Crypto Clipper and Card Skimmer on 178 Organizations - StepSecurity. Primary research blog. https://www.stepsecurity.io/blog/optimjack-optimizely-hosted-scripts-crypto-clipper-card-skimmer
- Shai-Hulud Descendant Surfaces in Malicious Ruby Gem That Steals Credentials and Poisons AI Tools | Socket. Primary research blog. https://socket.dev/blog/shai-hulud-descendant-malicious-ruby-gem
- Malicious bee-js npm Release and Swarm Desktop Releases Deliver SleepyDuck Ethereum C2 Backdoor | Socket. Primary research blog. https://socket.dev/blog/bee-js-compromise
- Socket raises $60M Series C at $1B valuation led by Thrive Capital to secure AI-driven software development | Socket. Primary research blog. https://socket.dev/blog/series-c
- Source referenced in coverage https://socket.dev/blog/optimizely-supply-chain-attack
SecurityXP delivers cybersecurity news, vulnerability analysis, data breach reports, and threat intelligence.
Security Digest
The most important cybersecurity stories, CVE alerts, and threat intelligence — one concise email. Unsubscribe anytime.
Related Articles
Malicious Microsoft agent-framework Releases on PyPI Run a Credential Stealer on Every Python Start
Malicious versions of the Microsoft agent-framework PyPI package shipped a credential stealer that runs on every Python start, researchers found.
Cold Relic Reused Stolen Google Workspace Data to Breach a Second U.S. Organization
Google says Cold Relic reused stolen identities and Workspace data from an earlier breach to compromise a second U.S. organization.
Cold Relic Reused Stolen Google Workspace Data to Breach a Second U.S. Organization
Google says Cold Relic reused stolen identities and Workspace data from an earlier breach to compromise a second U.S. organization.
ESLint's Payout Program Ended Over AI-Generated Code
ESLint ended its paid contributor program after AI-generated pull requests overwhelmed volunteer review capacity — a supply-chain integrity signal with downstream consequences.