Following an investigation, the methods of hacker group WP-SHELLSTORM have been discovered. They regularly attack WordPress websites.
A security investigation has provided insight into the methods of a cybercriminal group that compromises websites on a large scale. Because one of the group’s internal servers was left unsecured online for 22 days, researchers were able to analyze exploit tools, webshells, targets, and stolen cloud data.
Researchers from SOCRadar discovered that the group, named WP-SHELLSTORM, had left an unsecured Python server containing approximately 800 MB of files. These included exploit software, log files, command history, and thousands of webshells used to take over vulnerable websites. According to the researchers, this is a financially motivated cybercriminal operation rather than a state-sponsored attack.
Thousands of websites infected via known vulnerabilities
The leaked data shows that WP-SHELLSTORM did not use zero-days, but instead employed automated exploitation of known vulnerabilities in outdated WordPress plugins and Joomla components. The attack targeted 27 known security flaws, some of which were responsible for the majority of successful compromises.
The most frequently exploited vulnerability was found in the WordPress cache plugin Breeze (CVE-2026-3844). According to the log files, the attackers attempted to exploit this vulnerability on more than 45,000 websites. Ultimately, more than 17,000 webshells were installed, making this one of the largest known WordPress campaigns of the past year.
Although the infrastructure contained references to more than 1.4 million websites, these were largely scanned targets. Researchers eventually confirmed approximately 25,000 actually compromised websites, of which over 5,700 still contained an active webshell during the investigation.
Persistent access and stolen cloud data
After a successful breach, the attackers installed a hidden webshell that allowed them to execute commands remotely, view files, steal passwords, and move further within the victims’ infrastructure. For long-term access, an additional backdoor was also installed that disguised itself as a legitimate kernel process on Linux systems.
The open server also revealed a previous attack campaign. This showed that the group attacked vulnerable Nacos configuration servers to seize cloud credentials. Among the stolen data were access credentials for AWS, Oracle Cloud, Alibaba Cloud, Tencent Cloud, and DigitalOcean, alongside database passwords and cryptographic keys.
According to SOCRadar, this indicates that the group initially collected corporate credentials and later shifted its focus to compromising websites on a large scale.
