For most of Monday, visitors to some of the most heavily trafficked sites on the web were served a script that came from a domain their browsers had every reason to trust. According to monitoring by security researcher Scott Helme, malicious JavaScript was delivered through Optimizely Web, the A/B testing and experimentation platform, to visitors of roughly 178 organizations, spread across 188 compromised script instances. Later tallies circulating among researchers put the number of affected sites at about 186.
The injection point was not a forgotten plugin on a single site. Attackers wrote the payload into the Project JavaScript field that customers control inside the Optimizely dashboard, a slot meant for custom experiments. Because that code is served from Optimizely own content delivery network, every affected site inherited the same trusted origin. The campaign has been tagged OPTIMJACK.
The payload was a multi-purpose kit rather than a single trick. It replaced cryptocurrency wallet addresses copied to the clipboard with the attacker own address, harvested payment card details entered at checkout, and pulled cookies and screenshots from the browser. On iPhones and iPads it went further, delivering the Coruna and DarkSword exploit kits. Helme warned that the campaign was still active as of Monday evening.
The victim list reads like a cross-section of the Fortune 500 and the global media: Zoom, the BBC, CNN, Fox News, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger, Peacock, American Eagle and Hootsuite were all named, and Optimizely own website was affected as well. Not every listed organization has confirmed the exposure, and the total financial damage is not yet known.
The cash evidence is modest so far. By 19:25 UTC on Nov 16, roughly USD 7,000 in crypto had arrived in a wallet tied to the campaign. That figure measures only one chain and one collection address, so it is a floor rather than a total, and card data stolen on the same day typically surfaces in fraud losses weeks later.
What makes the incident hard to defend against is that nothing on the victim side had to be broken. Content security policies that allowlist a vendor CDN will wave the malicious script through, because the request goes to the vendor domain that the policy explicitly permits. Helme argues the practical answer is change detection rather than static allowlists: alerting when a trusted script changes, when a new script appears, or when an unexpected external destination shows up in the page, plus a policy in enforcement mode that actually blocks the risky calls.
The timing is uncomfortable for the wider industry. October already brought critical flaws in PHP and the Node.js runtime alongside a security incident at LiteLLM, the open-source proxy used to route requests between AI models, a cluster that some vendors now call Black October. In each case the target was shared infrastructure that thousands of downstream sites depend on and rarely audit.
The takeaway for security teams is unglamorous: an inventory of every third-party script on the estate, including marketing and experimentation tools that never pass through a code review, and a way to detect the moment one of those scripts changes shape. A tag manager or testing platform is a convenience layer until it becomes a distribution channel, and on Monday it distributed a card skimmer to a few hundred of the largest brands on the internet.
Comments (0)
Log in to join the discussion
Log InNo comments yet