Phishing with Cloudflare Workers: Transparent Phishing and HTML Smuggling

Phishing campaigns are abusing Cloudflare workers in two ways, where one injects malicious content hidden by HTML smuggling, similar to Azorult malware. 

The other uses Cloudflare Workers as a transparent proxy to steal login credentials for services like Microsoft, Gmail, and Yahoo Mail, which target users in Asia, North America, and Southern Europe, particularly in the tech, finance, and banking sectors. 

The rise in domains and applications hosting the attacks suggests continuous efforts by attackers to evade detection and takedowns, which highlights the common practice of abusing free cloud services for phishing, and the effectiveness of targeting popular cloud platforms. 

Attackers are abusing Cloudflare Workers, a free serverless platform for deploying applications, to host phishing sites, which allows them to create malicious applications with custom domains under the workers.dev subdomain and distribute them freely. 

 All-in-One Cybersecurity Platform for MSPs to provide full breach protection with a single tool, Watch a Full Demo

While the overall traffic volume seems to have stabilized, the number of distinct malicious applications hosted on Cloudflare Workers continuously grows, indicating a concerning trend of attackers adopting this technique. 

Attackers are using HTML smuggling with Cloudflare Workers to bypass network defenses and deliver phishing pages.

Malicious code is hidden within a seemingly harmless webpage as a base64-encoded blob, further obfuscated through multiple encodings. 

The createObjectURL() method makes the hidden phishing page accessible within the endpoint, and the click is simulated on the blob URL to display the phishing page in the victim’s browser.  

They use transparent phishing, a new technique that bypasses traditional phishing limitations and works as the attacker creates a server (often hosted on Cloudflare Workers) that acts as a middleman between the victim and the legitimate login page. 

The victim sees the real login page but unknowingly submits their credentials (including multi-factor codes) to the attacker’s server, which then forwards them to the real service.  

Researchers at Netskope analyzed these attacks and found they’re built on modified open-source MITM toolkits and leverage Cloudflare Workers’ “Hello World” template to intercept victim requests. 

The attacker’s application acts as a transparent proxy, first forwarding the victim’s request to the legitimate site and capturing the response. 

It then replaces the legitimate site’s domain with its own before displaying the response to the victim, which deceives the victim into entering their credentials on the attacker’s application. 

Upon login, the attacker’s application steals the tokens and cookies from the legitimate site’s response and gains visibility into the victim’s subsequent activity. 

Get special offers from ANY.RUN Sandbox. Until May 31, get 6 months of free service or extra licenses. Sign up for free.