Dead Domain, Live Danger: The Ghost CDN Quietly Controlling Thousands of Sites
Friday, September 18, 2026 · 5-minute read
Researchers found 13 malicious npm packages delivering a new stealer called WeaselBiscuit. The malware shares code with tools previously tied to North Korea's "Contagious Interview" hacking group. That campaign typically targets software developers by posing as job recruiters, then slipping malware onto their machines during a fake hiring process.
A separate piece of malware called PhantomRaven also appeared on npm — and researchers believe a financially motivated criminal wrote it using an AI chatbot, based on the oddly verbose comments baked into the code.
↗ The Hacker NewsThe FBI took down NightmareStresser, a service that let anyone pay to flood a website or server with traffic until it went offline. These DDoS-for-hire platforms are sometimes called "booters" and are commonly used to knock gaming servers and small businesses offline. The seizure removes one of the most established platforms of its kind from the market.
↗ BleepingComputerA New Jersey judge ordered the domain radaris.com transferred to the plaintiffs after the company's owners — Russian-born brothers Igor and Dmitry Lubarsky — repeatedly dodged legal proceedings using a rotating cast of shell companies in places like the Marshall Islands and the British Virgin Islands. The lawsuit was brought by Atlas Data Privacy Corp under New Jersey's Daniels Law, which protects law enforcement officers and government employees from having their personal information published online. It's a rare win against a data broker that spent years winning by exhausting its opponents.
↗ Krebs on SecurityAttackers stole a Cloudflare API key from Brevo, an email and marketing platform. They used it to inject malicious ClickFix scripts into Brevo's own websites and into JavaScript files that Brevo customers embed on their own sites. Anyone who visited an affected customer site could have been exposed to malware. This is a supply-chain attack — instead of hacking thousands of sites individually, the attackers hit one supplier and got access to all of them at once.
A new Android malware called RatHat uses an AI-powered subsystem to help attackers remotely navigate an infected phone — tapping buttons, opening apps, and reading screens. Unusually, it abuses ADB (Android Debug Bridge) to keep a shell connection open even after the malicious app is uninstalled. That means simply deleting the app may not be enough to evict it.
A maximum-severity flaw in Microsoft's Azure AI Foundry platform — the tool enterprises use to build and deploy AI apps — let an attacker gain elevated access to systems without any login credentials at all. A missing authentication check on a critical function meant anyone on the network could just walk in and escalate their own privileges. A privilege escalation flaw this severe on a cloud AI platform used by large organizations is about as serious as it gets.
Status: Patch already applied by Microsoft — no action required from customers. No evidence of exploitation in the wild.
Cisco has warned of a maximum-severity zero-day in its Identity Services Engine — software that many organizations use to control who gets onto their corporate network. The flaw lets attackers execute code with root privileges, meaning full control of the system. Critically, this one is already being exploited in real attacks.
Status: Security updates released by Cisco. If your organization runs Cisco ISE, apply patches immediately.
Check Point — a major cybersecurity vendor — released emergency updates for a flaw in its own security management systems. The vulnerability lets attackers execute code with root privileges, which is the highest level of system access possible. It's a reminder that security tools themselves can have serious holes.
Status: Security updates available from Check Point. Apply them now if your organization uses Check Point management systems.
Security firm Air Security revealed a flaw it calls "Plugin4Shell" affecting four widely used AI coding agents: Claude Code, OpenAI Codex, GitHub Copilot, and Google's Gemini CLI. When these agents install plugins, they're supposed to lock each plugin to an exact reviewed version using a commit hash. The problem: none of them actually verify the code they download matches that fingerprint. A malicious repository owner could create a cleverly named branch that tricks the agent into downloading different — potentially harmful — code instead. Anthropic patched the flaw in Claude Code 2.1.179 and OpenAI fixed it in Codex 0.146.0. GitHub Copilot has no fix yet, and Google won't patch the Gemini CLI because it's being retired.
Microsoft is rolling out a new Teams feature that lets company administrators customize which file types employees can share through the platform. Currently, Teams blocks a default list of extensions commonly associated with malware — like .exe files. The new control lets admins add their own risky file types to that list, reducing the risk of someone accidentally receiving and opening a malicious file through a chat.
Imagine a plumber installs pipes in your house connected to a water company that later goes out of business. The company's infrastructure disappears — but your pipes are still connected to their old supply line. Now a stranger buys that supply line and decides what flows through your taps. That's essentially what happened here, and thousands of websites are currently in that situation right now.
In July 2025, someone registered a domain that previously belonged to a defunct CDN (content delivery network). The CDN had shut down years earlier and let the domain expire. But hundreds of websites, code repositories, and documentation pages still had hard-coded references to files hosted on that domain — things like fonts, scripts, and images. When the new owner registered the domain and set up wildcard DNS, every single one of those requests started going to infrastructure they control. Today the domain serves an ad-heavy downloader page. Tomorrow it could serve malware — and nothing would look broken from the outside.
This isn't theoretical. In June 2024, the polyfill.io domain — a piece of JavaScript embedded in more than 110,000 websites — changed hands and began redirecting mobile visitors to malicious sites. The new owners had simply bought a trusted name and weaponized it. The scary part of the CDN story reported today is that nobody has been notified. The thousands of websites still calling that domain have no idea their content now comes from a stranger's server.
For regular website owners, the lesson is this: audit your site for any hard-coded references to external scripts and resources — especially ones pointing to domains you don't control. If you're not sure where your site loads files from, ask your web developer to run a check. A dormant reference you forgot about years ago could become a live problem the moment someone else buys that domain.