HIGHLIGHT OF THE DAY · CYBERSECURITY
Optimizely confirms script attack; cleaned files leave visitor harm unmeasured
Optimizely says it fixed a vulnerability used to distribute malicious website scripts. Researchers found poisoned snippets tied to 178 organizations and later observed cleanup, but the number of visitors who suffered theft or device compromise remains unknown.
· 5 min read · 8 sources
Briefing24 · AI-assisted research and analysis · Methodology
Vendor confirms exploitation of Web Experimentation
Optimizely confirmed on November 17 that attackers exploited its Web Experimentation service, affecting snippet delivery. Its incident notice says the vulnerability was fixed and injected code removed, but the status remains Monitoring. In a statement obtained by BleepingComputer, the company identified the vulnerable component as its Experimentation account-signup flow. It said it blocked that path, ended sessions, revoked tokens, reset affected account access and notified affected customers. [2][8]
Report URI examined 1,611 distinct snippets and found 188 malicious ones belonging to 178 organizations, including BBC, CNN, Zoom and Salesforce. Those figures describe a research sample, not a complete customer or visitor census. The distinction matters for named organizations: Hudson Rock documented GitLab’s marketing website and HSBC’s UAE public website, not access to source repositories, banking systems or customer workspaces. [3][4]
Project configuration became the distribution channel
The attackers placed their loader in customers’ Project JavaScript configuration. Optimizely’s documentation says this code runs when the snippet loads, before subsequent experiment-selection steps. Together, the documentation and recovered scripts explain how the legitimate delivery mechanism could carry attacker-controlled code without a change to a customer’s own application repository. Report URI identifies the incident as poisoned project content, not CDN cache poisoning. [3][5]
Researchers recovered plain, Base64-obfuscated and embedded-payload variants. Some included an inline fallback, so blocking the external payload domain alone could not reliably neutralize every poisoned snippet. This makes inspection of project content a necessary counterpart to network controls, rather than treating a blocked destination as proof that the malicious code was removed. [3]
Corroboration extends beyond repeated headlines. Cynet recorded endpoint alerts and browser captures involving CNN and Zoom, while BleepingComputer independently examined modified files and archived page scans. StepSecurity separately checked payloads, DNS and wallets. These provide distinct evidence of delivery and behavior. By contrast, SecurityWeek’s repetition of Report URI’s organization count is not a second independent census. [1][6][7][8]
Cleanup is documented, but exposure boundaries are not
Report URI’s first observed injection was in the BBC snippet at 04:21 UTC on November 16. One of its 188 identified snippets still contained malicious code at 21:52. Its subsequent update said all identified infected files were clean by 23:11 and remained clean in its observations. Those timestamps describe the researcher’s evidence, not the actual beginning of the attack or a vendor-certified exposure window for every customer. [3]
StepSecurity reported that the main payload domain, 1688360[.]cc, stopped resolving after being placed on hold at 00:37 UTC on November 17, although some associated infrastructure still resolved at 07:30. Optimizely published its confirmation at 01:02. The synthesis supports substantial containment, not a claim that malicious delivery remains universally active. Our inference is that historical exposure still requires investigation: cleaned files cannot establish whether earlier visitors suffered harm. [2][3][6]
Financial-theft capabilities had browser-level limits
StepSecurity’s analysis describes code capable of substituting cryptocurrency addresses in page content, form values, clipboard interactions and requests. It also found payment-card skimming logic for accessible card numbers, expiry dates and security codes. Cross-origin hosted payment fields were outside that skimmer’s reach, so a finding on a public webpage does not by itself establish that card details entered elsewhere were accessible. [6]
Cookie and page-image collection was conditional, triggered after an address swap, rather than indiscriminate collection from every visitor. HttpOnly cookies were inaccessible to the JavaScript. Cynet also cautions that rendering a webpage is not a full-desktop screenshot and that browser-level code does not automatically reach unrelated tabs, origins or sessions. A capability in code, an outbound request and verified sensitive-data theft remain different findings. [6][7]
Mobile loaders and wallet balances leave harm unproven
Hudson Rock linked a mobile delivery branch to Coruna- and DarkSword-associated infrastructure, but did not demonstrate successful exploitation of a real iPhone. Some tests used Chrome with an iOS user-agent string, which can expose conditional delivery without reproducing Safari/WebKit or the iOS kernel. Cynet independently retrieved a secondary mobile loader with an emulated profile. That corroborates delivery, not successful physical-device compromise, and the router’s version branches are not a tested vulnerability matrix. [4][7]
Financial-loss claims also need qualification. Report URI described approximately $7,040 received by attacker-associated wallets as stolen funds. StepSecurity found that some addresses had activity predating this incident and did not independently reproduce the entire TRON total. Wallet receipts therefore cannot be treated as a verified incident-specific loss figure. Counts of stolen cards, exposed cookies and compromised devices remain undetermined. [3][6]
Operators should preserve evidence and inspect project content
Website operators should review every Project JavaScript field and experiments’ custom code for unauthorized additions, including encoded and inline variants. They should audit users, API tokens and integrations with write access, and rotate affected credentials. Available snippets, timestamps and browser or network evidence should be preserved to reconstruct what visitors could have received, rather than relying only on today’s clean response. [3][7]
Sensitive pages need separate assessment. A public homepage finding does not establish that checkout, online banking or account portals loaded the same payload. Conversely, a clean response now does not invalidate an earlier malicious capture. Blocking validated malicious destinations is useful, but Cynet’s observations do not justify classifying the entire legitimate Optimizely CDN as malicious. [4][7]
Potentially exposed visitors can review cryptocurrency destinations used during the relevant period, monitor card activity and contact their issuer about suspicious transactions. Updating iPhones and iPads to the latest supported software is also a precaution. These steps do not imply that every visitor’s account or device was compromised. [8]
The next disclosures need to connect delivery to outcomes
The outstanding questions are project-level modification timelines, how the signup flaw enabled changes across unrelated customer projects, independently validated exposure totals and evidence of completed theft or device exploitation. Optimizely’s public advisory supplies neither a CVE nor a customer-version remediation matrix. A formal resolution would clarify containment, but our assessment is that measuring harm requires additional evidence beyond service status and cleaned snippets. For now, confirmed malicious-script distribution should not be presented as mass compromise of every named company or visitor. [2][3][4][6][8]
Sources & further reading
- Optimizely Vulnerability Exploited to Serve Malware on Major Websites - SecurityWeek ↗www.securityweek.com
- Optimizely Service Status - Web Experimentation Security Incident ↗status.optimizely.com
- World's Largest Brands Infected With Malware Via Optimizely ↗blog.report-uri.com
- Poisoned Scripts on BBC, Zoom and Other Major Websites Linked to iPhone Exploit Loaders | InfoStealers ↗www.infostealers.com
- Snippet order of activation - Web Experimentation docs ↗docs.optimizely.com
- OPTIMJACK: Optimizely-Hosted Scripts Loaded a Crypto Clipper and Card Skimmer on 178 Organizations - StepSecurity ↗www.stepsecurity.io
- OPTIMJACK: Malicious web injection across CNN, Zoom, Salesforce, Fox and more - Cynet ↗www.cynet.com
- Optimizely hack infected CNN, Fox News, BBC, Salesforce sites with malware ↗www.bleepingcomputer.com
Researched, written and checked with GPT-6 Astra. Publication is automatic after source, structure and model review checks. These checks can miss errors and do not constitute human verification.
Report a correction · Browse highlights · Read the daily briefing