React2Shell flaw exploited to breach 30 orgs, 77k IP addresses weak

Over 77,000 Web-exposed IP addresses are weak to the essential React2Shell distant code execution flaw (CVE-2025-55182), with researchers now confirming that attackers have already compromised over 30 organizations throughout a number of sectors.

React2Shell is an unauthenticated distant code execution vulnerability that may be exploited by way of a single HTTP request and impacts all frameworks that implement React Server Parts, together with Subsequent.js, which makes use of the identical deserialization logic.

React disclosed the vulnerability on December 3, explaining that unsafe deserialization of client-controlled knowledge inside React Server Parts allows attackers to set off distant, unauthenticated execution of arbitrary instructions.

Builders are required to replace React to the newest model, rebuild their purposes, after which redeploy to repair the vulnerability.

On December 4, safety researcher Maple3142 printed a working proof-of-concept demonstrating distant command execution towards unpatched servers. Quickly after, scanning for the flaw accelerated as attackers and researchers started utilizing the general public exploit with automated instruments.

Over 77,000 weak IP addresses

Shadowserver Web watchdog group now experiences that it has detected 77,664 IP addresses weak to the React2Shell flaw, with roughly 23,700 in the US.

The researchers decided that IP addresses have been weak utilizing a detection method developed by Searchlight Cyber/Assetnote, the place an HTTP request was despatched to servers to use the flaw, and a particular response was checked to substantiate whether or not a tool was weak. 

GreyNoise additionally recorded 181 distinct IP addresses trying to use the flaw over the previous 24 hours, with a lot of the site visitors showing automated. The researchers say the scans are primarily originating from the Netherlands, China, the US, Hong Kong, and a small variety of different nations.

Palo Alto Networks experiences that greater than 30 organizations have already been compromised via the React2Shell flaw, with attackers exploiting the vulnerability to run instructions, conduct reconnaissance, and try to steal AWS configuration and credential information.

These compromises embrace intrusions linked to identified state-associated Chinese language menace actors.

Widespread exploitation of React2Shell

Since its disclosure, researchers and menace intelligence corporations have noticed widespread exploitation of the CVE-2025-55182 flaw.

GreyNoise experiences that attackers incessantly start with PowerShell instructions that carry out a fundamental math perform to substantiate the gadget is weak to the distant code execution flaw.

These exams return predictable outcomes whereas leaving minimal indicators of exploitation:

powershell -c "40138*41979"
powershell -c "40320*43488"

As soon as distant code execution was confirmed, attackers have been seen executing base64-encoded PowerShell instructions that obtain extra scripts immediately into reminiscence. 

powershell -enc 

One noticed command executes a second-stage PowerShell script from the exterior website (23[.]235[.]188[.]3), which is used to disable AMSI to bypass endpoint safety and deploy extra payloads.

In keeping with VirusTotal, the PowerShell script noticed by GreyNoise installs a Cobalt Strike beacon on the focused gadget, giving menace actors a foothold on the community.

Amazon AWS menace intelligence groups additionally noticed fast exploitation hours after the disclosure of the React CVE-2025-55182 flaw, with infrastructure related to China-linked APT hacking teams often called Earth Lamia and Jackpot Panda. 

On this exploitation, the menace actors carry out reconnaissance on weak servers through the use of instructions corresponding to whoami and id, trying to jot down information, and studying /and many others/passwd.

Palo Alto Networks additionally noticed related exploitation, attributing a few of it to UNC5174, a Chinese language state-sponsored menace actor believed to be tied to the Chinese language Ministry of State Safety.

“Unit 42 noticed menace exercise we assess with excessive confidence is in step with CL-STA-1015 (aka UNC5174), a bunch suspected to be an preliminary entry dealer with ties to the Chinese language Ministry of State Safety,” Justin Moore, Senior Supervisor at Palo Alto Networks Unit 42, advised BleepingComputer by way of e mail.

“On this exercise, we noticed the deployment of Snowlight and Vshell malware, each extremely in step with Unit 42 information of CL-STA-1015 (often known as UNC5174).”

The deployed malware in these assaults is:

• Snowlight: A malware dropper that permits distant attackers to drop extra payloads on breached gadgets.
• Vshell: A backdoor generally utilized by Chinese language hacking teams for distant entry, post-exploitation exercise, and to maneuver laterally via a compromised community.

The push to patch

Because of the severity of the React flaw, corporations worldwide have rushed to put in the patch and apply mitigations.

Yesterday, Cloudflare rolled out emergency detections and mitigations for the React flaw in its Net Utility Firewall (WAF) as a result of its widespread exploitation and severity.

Nevertheless, the replace inadvertently brought about an outage affecting quite a few web sites earlier than the principles have been corrected. 

CISA has additionally added CVE-2025-55182 to the Identified Exploited Vulnerabilities (KEV) catalog, requiring federal businesses to use patches by December 26, 2025, beneath Binding Operational Directive 22-01.

Organizations utilizing React Server Parts or frameworks constructed on prime of them are suggested to use updates instantly, rebuild and redeploy their purposes, and evaluation logs for indicators of PowerShell or shell command execution.

Damaged IAM is not simply an IT downside – the affect ripples throughout your entire enterprise.

This sensible information covers why conventional IAM practices fail to maintain up with fashionable calls for, examples of what “good” IAM appears like, and a easy guidelines for constructing a scalable technique.

Get the information