Hackers have breached 25 servers, including three linked to the US government, by chaining the critical PHP2Shell vulnerability with a Linux kernel privilege escalation flaw to gain root access, according to researchers at Dream Security. The US Cybersecurity and Infrastructure Security Agency has issued a warning about CVE-2026-46300, known as Fragnesia, which allows attackers who first exploit PHP2Shell bugs to run malicious code with root permissions. Once attackers secure root access, they gain unrestricted control over servers, including the ability to disable systems, create or remove accounts, connect to other servers, and install or delete software.

The Dream researchers tracked an attack in which threat actors constructed a persistent tunnel into an internal network within hours, developing two exploits that targeted any Palo Alto Networks firewall running a vulnerable version of PAN-OS. After gaining root access through a single HTTP request without credentials, the attackers launched strikes against Active Directory, extracting credentials and discovering hidden files, folders, and other sensitive resources. The threat actor obtained a complete AD user group listing and organizational blueprint, private keys and full network topology with IP addresses, hostnames, server roles, and VPN gateways, plus decrypted service account credentials enabling direct AD queries at any time. The researchers confirmed that the adversarial actor also parked nearly two dozen firewalls across the US, Japan, Taiwan, and South Korea, ready for exploitation at any time.

PHP2Shell allows remote access, but that access is typically limited by the account used to run the PHP service, explained Scott Miserendino, CTO at DataBee. "By chaining with CVE-2026-46300, that attacker can get unlimited access," he said. The two PHP bugs gave attackers code execution privileges using the web service account, and the kernel bug is what turned that into root, noted Aaron Beardslee, manager of threat research at Securonix, who pointed out that the kernel bug writes its payload into the page cache rather than to disk. The breach did not appear to be a fully autonomous nation-state attack but did include some AI-assisted tooling like Claude Code and AI-powered offensive security platforms configured with Chinese large language models DeepSeek and Kimi, according to the Dream researchers.

The privilege escalation step in this chain is invisible to disk-based integrity checking, and the standard remediation advice destroys the evidence of it, Beardslee explained. The only remedy for CVE-2026-46300 is to upgrade to a patched Linux version, according to Miserendino, though options short of full patching could break the system or require IPSec encapsulating security payload to be disabled, potentially introducing additional flaws. Beardslee recommended patching the kernel and capturing memory first on anything suspected of compromise, noting that the Fragnesia fix is not included in the Dirty Frag patches and that the reboot is what erases the evidence. He also advised auditing the core built-in kernel modules esp4 and esp6 on every internet-facing Linux host, and checking which hosts hold a directory service credential, then alerting on LDAP binds from anything that isn't on that list to catch the chain at the stage where it stops being a web server problem and starts being a domain problem. Organizations that rely on perimeter defenses alone may find that exposed management interfaces now represent a direct path to their most sensitive identity infrastructure, forcing security teams to rethink where they monitor for compromise and how quickly they can respond before attackers pivot from initial access to domain control.