Retour 
Summary
Netskope Threat Labs is tracking multiple phishing campaigns that abuse Cloudflare Workers. The campaigns are likely the work of different attackers since they use two very different techniques. One campaign (similar to the previously disclosed Azorult campaign) uses HTML smuggling, a detection evasion technique often used for downloading malware, to hide the phishing content from network inspection. The other uses a method called transparent phishing, where the attacker uses Cloudflare Workers to act as a reverse proxy server for a legitimate login page, intercepting traffic between the victim and the login page to capture credentials, cookies, and tokens.
Netskope Threat Labs has been tracking an increasing number of Netskope users targeted by malicious content hosted in Cloudflare Workers throughout 2023 and into 2024. The number of targeted users appears to have leveled off so far in 2024, although the number of domains continues to increase. At the same time, the distinct number of applications hosting the malicious content continues to increase, indicating that attackers are constantly creating new apps to evade detections and takedowns.
Over the past 30 days, phishing campaigns hosted on Cloudflare Workers have primarily targeted victims in Asia, North America, and Southern Europe, across multiple segments led by technology, financial services, and banking. The majority of the phishing pages Netskope Threat Labs has uncovered target Microsoft login credentials, with Gmail, Yahoo Mail, and cPanel Webmail among the other targets.
Abusing free cloud services to host malicious content, including malware and phishing pages, is a common practice among adversaries. Our most recent Threat Stats post highlights that no cloud apps are immune to such abuse and that abusive content hosted on the most popular apps tends to be most successful in reaching its victims.
Let’s take a closer look at these campaigns:
Cloudflare Workers serving phishing sites
Cloudflare Workers is a serverless computing platform for application deployment, including serving HTML pages to visitors. Cloudflare Workers is available to free tier users and therefore abused by attackers who continuously abuse free cloud services. With a free tier account, attackers can create multiple Cloudflare Worker applications that can serve up to 100,000 access a day from their victims, using a free publicly accessible domain and a valid TLS certificate. Netskope Threat Labs has previously written about the abuse of other free services, including Cloudflare R2.
Netskope Threat Labs first observed an increase in traffic to phishing pages hosted in Cloudflare Workers in Q2 2023 and spiking in Q4 2023. So far in 2024, the number of users targeted with malicious content hosted in Cloudflare Workers appears to have leveled off, but we are still seeing thousands of Netskope users attempting to access malicious content hosted in Cloudflare Workers each quarter.
At the same time, the number of distinct malicious applications that Netskope users are attempting to visit is growing. Each application has a distinct domain of the format https://{application-name}.workers.dev. The following graph shows the number of distinct domains to be steadily increasing throughout 2023 and into 2024 and has not yet leveled off.
Phishing pages smuggled through Cloudflare Workers
Several phishing campaigns hosted in Cloudflare Workers use HTML smuggling to deliver phishing pages to their victims. As described in our previous blog post, HTML smuggling is a defense evasion technique that attempts to bypass network controls by assembling the malicious payloads on the client side. In that post, the attacker saved the malicious payloads to the disk for the victim to execute. In this case, the malicious payloads are the phishing pages themselves, so the attacker simply reconstructs them and displays them in the browser. In both cases, the objective is to try to evade network-based defenses.
The attackers have embedded the actual phishing page as a blob inside a benign web page. The phishing page is initially encoded in base64 and then encoded multiple additional times to obfuscate the code and avoid static detection. To make the blob object accessible in the endpoint, they use the createObjectURL() method to create a blob URL. Then, they simulate a click on the blob URL using the click() method.
