Optimjack: attackers planted skimming code on 178 Optimizely customer websites
A flaw in Optimizely's signup flow let an attacker inject card-skimming, crypto-stealing JavaScript into Web Experimentation customer sites, including Zoom, the BBC, CNN and Salesforce.
Websites that load Optimizely's Web Experimentation snippet, a small piece of JavaScript the vendor serves to run A/B tests, delivered malicious code to visitors on 16 November 2026. An attacker abused a flaw in Optimizely's account signup flow to impersonate users on targeted customer accounts and drop their own JavaScript into those customers' projects, Optimizely said on its status page.
The tampered files came from Optimizely's own content delivery network under Optimizely's own certificate, so affected sites treated them as trusted. Researchers nicknamed the attack OPTIMJACK. The code swapped cryptocurrency wallet addresses in visitors' browsers, skimmed payment card details from checkout forms, and tried to push exploit kits at iPhones and iPads, according to Palo Alto Networks' Unit 42. Analysis by Report URI and BleepingComputer also found code to collect cookies, page URLs and screenshots.
Optimizely was not the only delivery channel. Hudson Rock found scripts from Kissmetrics, StorifyMe and Minute-ly serving a small loader that led to the same iOS-targeting infrastructure, with the earliest observation dated 13 November, SecurityWeek reported. Hudson Rock has confirmed links to the exploit-kit infrastructure but says successful device exploitation has not been demonstrated.
Optimizely says the incident is contained. It spotted the activity on the day it started, stopped delivery of the code, closed the vulnerability within three hours of detection, stripped the attacker's code out of every project it had reached, and says nothing malicious remains in its systems. It also says this was not a credentials breach, that it was limited to Web Experimentation, and that fewer than 5% of its customers were impacted.
Scott Helme's team at Report URI analysed 1,611 Optimizely scripts and found malicious code in 188 of them, belonging to 178 organisations, Dark Reading reported. BleepingComputer independently confirmed the script loading on sites including Zoom, KLM, GitLab, BBC, CNN, Smartsheet and Salesforce, with Fox News, Air France, Bitly, HSBC Insurance and UNICEF USA also named. Optimizely's own www and login snippets carried the code too, The Register reported. Thirty-two finance or banking sites were among those infected.
Salesforce told BleepingComputer it is investigating malicious modifications to Optimizely scripts on some of its websites and has disabled the scripts across preview and production instances as a precaution while it does so.
What to do now
If you do not use Optimizely Web Experimentation, you need take no action. Optimizely says no other product was affected.
If you do use it:
- Check whether Optimizely contacted you. The company says it reached out to every Web Experimentation customer directly to confirm whether they were affected and to supply remediation guidance. Keep an eye on the incident status page, which Sophos X-Ops also recommends monitoring.
- Open the Project JavaScript field for each of your projects and read it. That custom code field is where the attacker's loader was placed. Helme said infected files have all been cleaned since the incident was reported, but still urged companies using Optimizely to check their own sites.
- Do not rely on blocking the attacker's domain alone. Hudson Rock warned that some poisoned scripts carried a full embedded copy of the payload that runs if the remote file fails to load.
- Audit every change made to your projects while the malicious code was live, re-check any altered code, and treat payment data that passed through affected pages as potentially compromised, as Unit 42 advises.
- Work out which of your pages loaded the snippet and for how long. Black Duck's Collin Hogue-Spears told The Record that the skimmer reads forms a site built itself, while "a payment processor's hosted iframe sits outside its reach, so security teams at affected companies must reconstruct which pages loaded the infected Optimizely snippet".
- If you took card payments on pages using your own form fields, assume those details could have been captured and follow your card scheme's incident process. Where checkout uses a hosted iframe, a separate analysis by StepSecurity, reported by The Register, suggests those fields were out of reach, though Optimizely has not confirmed that.
- Review your other third-party tags. The same loader was seen in Kissmetrics, StorifyMe and Minute-ly scripts from 13 November.
Advice for people who visited affected sites, per BleepingComputer: review any cryptocurrency transfers made on 16 November to confirm the destination address was not swapped, watch card statements for unauthorised transactions, update iPhones and iPads to the latest iOS or iPadOS, and keep an eye on accounts you were logged into at the time.
Longer-term hardening. Unit 42 suggests stripping non-essential marketing, experimentation and analytics tags from login and checkout pages, and isolating payment collection in hardened cross-origin iframes. Fortra's Tyler Reguly pointed to Subresource Integrity (SRI), a browser feature that checks an external script has not been altered before running it, though Optimizely's documentation says the snippet updates automatically to run new experiments. Vigilant's Chris Nyhuis argued the answer is not to drop third-party tools but to restrict where vendor scripts run, limit their access to sensitive data and monitor their behaviour.
Technical details
No CVE identifier has been cited in reporting so far. Optimizely attributes the incident to a vulnerability in the Web Experimentation account signup flow, used by an unauthorised third party to impersonate users on targeted customer accounts and insert JavaScript into their projects. The affected component on the status page is listed as Web Experimentation (Snippet Delivery).
Timeline: activity began on 16 November 2026 and was detected the same day, with delivery blocked and the flaw closed within three hours. Optimizely published an initial update on 17 November 2026 and a fuller account on 18 November 2026. Helme's data showed the earliest observed Optimizely injection in the early morning of 16 November, though how long the vulnerability existed and how many visitors were exposed remain unknown. Helme also noted the injected code appeared on Turkish websites in September, before the larger campaign. By 21:52 UTC on 16 November, only one of the 188 compromised files still contained the code, and The Register reported that site as TalkTalk.
Indicators: the injected loader pulled a second script from https://1688360[.]cc/track.js, and captured card data was sent to https://1688360[.]cc/card.php. Unit 42 found three loading variants: a direct URL reference, a Base64-encoded URL decoded with atob(), and a fallback that executes an embedded copy if the remote script is blocked. Sophos lists the detection name Troj/JSInj-H. The iOS payloads were the Coruna and DarkSword exploit kits; The Record notes DarkSword can infect visitors with no click, while Coruna arrives via web attacks hitting WebKit and kernel flaws. iVerify has warned that about 75% of the spyware detections it tracks involve one of the two kits.
Helme tracked roughly $7,000 received by wallets embedded in the script as of 16 November, though some were active before the attack, so the true figure is unclear. He said it is impossible to say "how many credit cards were skimmed, how many cookies or screenshots were stolen, or how many iOS devices were infected before the attack was stopped".
Optimizely's remediation, per a spokesperson quoted by Dark Reading, included "blocking access to the vulnerable path, deploying a fix, removing the injected code from affected projects, ending active sessions, revoking tokens, and resetting access for affected user accounts", plus extra protections in Cloudflare. The company says the incident is fully remediated while an investigation with third-party experts continues. It also disclosed a separate incident in February, when a threat actor reached its systems through a voice-phishing attack.
The central question is still unanswered: how a flaw in a signup flow granted write access across unrelated customer tenants. Report URI said it could not determine this, and Optimizely did not answer Dark Reading's question on the point. The Media Trust is tracking the same malware under the name Hydra-Cart and describes the threat as contained.
Share this article
What changed
- : Added that the same iOS loader was served via Kissmetrics, StorifyMe and Minute-ly scripts from 13 November, that Salesforce disabled Optimizely scripts, and Hudson Rock's fallback payload warning.