3 Sources
[1]
PHP2Shell Attack Roots 25 Servers, Breaches U.S. Government-Affiliated Network
Cybersecurity researchers have disclosed details of what appears to be an artificial intelligence (AI)-assisted attack orchestrated by a suspected China-nexus threat actor that has leveraged the PHP2Shell vulnerabilities to breach exposed systems. The activity, which was observed on October 25,
[2]
PHP2Shell Exploitation Hits Government Entity Amid Global Scanning and Signs of AI-Assisted Hacking
Threat actors have been observed using agentic artificial intelligence (AI) platforms for post-exploitation activity following the abuse of the PHP2Shell vulnerabilities tracked as CVE-2026-109618 and CVE-2026-109619. "In one case, the attackers targeted a government entity in Europe and within
[3]
FBI urges immediate PHP patching as exploit attempts surge
The FBI urged immediate security updates for two critical PHP vulnerabilities discovered using artificial intelligence, as governments and companies patched systems and attackers launched millions of exploit attempts. The flaws, CVE-2026-109618 and CVE-2026-109619, are collectively known as
Share
Copy Link
A suspected China-nexus threat actor leveraged AI tools to exploit PHP2Shell vulnerabilities, rooting 25 servers across multiple countries including three systems in a U.S. government-affiliated environment. The FBI and CISA have issued urgent warnings as millions of exploit attempts flood vulnerable systems worldwide.
Cybersecurity researchers have uncovered an AI-assisted cyberattack that exploited PHP2Shell vulnerabilities to compromise 25 servers across the U.S., Japan, Taiwan, and South Korea, including three systems within a U.S. government-affiliated network
1
. The attack, observed on October 25, 2026, combined two critical PHP vulnerabilities—CVE-2026-109618 and CVE-2026-109619—with a Linux kernel privilege escalation flaw tracked as CVE-2026-46300 to achieve unauthenticated remote code execution and root access1
.The threat actor, suspected to have China-nexus ties, deployed custom exploit code that differed from publicly available versions. Dream, the cybersecurity firm that discovered the activity, noted that the implementation "appears to have been independently developed and tailored specifically for the targeted firewall platform"
1
. The affected devices ran specific versions of a large vendor's firewall software using PHP 8.0.30, though the vendor's name was not publicly disclosed.Artifacts recovered from the attack infrastructure reveal the use of AI-powered tools throughout the operation. Researchers discovered installations of Anthropic's Claude Code, API keys for the FOFA search engine, and an AI-powered offensive security platform called aiscan configured with large language models from DeepSeek and Kimi
1
. The code structure, commenting style, and consistency across nearly 40 scripts strongly suggest AI augmentation during development.In a separate incident targeting a government entity in Europe, Palo Alto Networks Unit 42 observed attackers using agentic artificial intelligence platforms for post-exploitation activity. Within two hours of initial exploitation, the attackers successfully accessed password hashes and multi-factor authentication secrets, established an NTLM-relay tunnel into Active Directory, and attempted Domain Admin privilege escalation
2
. Commands were wrapped in section-header delimiters like "==SHELLUSER==" and "==PASSWD==" to produce machine-parseable output for AI model consumption2
.The attack unfolded through a methodical process that demonstrates sophisticated vulnerability management and targeting. The threat actor used FOFA to identify approximately 19,000 internet-facing firewalls with exposed management interfaces worldwide
1
. A custom version scanner then determined which devices were online and their operating system versions, with about 5,000 devices responding with version information.The attacker filtered results to systems in the U.S., Japan, Taiwan, and South Korea running PHP 8.0.30, yielding 240 targets. An automated batch runner exploited the PHP2Shell-Fragnesia chain, planted PHP web shells on publicly accessible directories, verified root access, and logged target information to a CSV file—successfully compromising 24 servers
1
.Two compromised systems belonged to a U.S. government-affiliated organization. One system crashed on October 23, 2026, at 6:14 p.m. UTC and never recovered. The attacker successfully rooted a second system on October 24, 2026, at 10:09 a.m. UTC, then built a persistent channel into the internal network via a reverse SSH tunnel
1
. By 11:35 a.m. UTC, a third government server—an internal firewall management server only reachable through the tunnel—was compromised using the same method.The attacker then targeted Active Directory to decrypt the LDAP service account password, enumerate users, and spray passwords against every user account using organization-themed passwords. The operation also attempted to create backdoor Active Directory accounts
1
. Whether the threat actor successfully extracted valid credentials or established backdoor accounts remains unclear, leaving the full extent of the compromise indeterminate.CYFIRMA uncovered a separate suspected North Korea-linked campaign employing PHP2Shell to probe global PHP infrastructure from 21 compromised nodes. The scanning targeted 216 unpatched hosts across the U.S., Japan, South Korea, Thailand, Indonesia, Australia, Taiwan, India, Singapore, the U.A.E., Saudi Arabia, and the E.U., spanning government, financial institutions, logistics, airports, manufacturing, and online transaction platforms
2
.CYFIRMA attributed this activity with moderate confidence to a Korean-speaking operator affiliated with the Lazarus Group ecosystem, coordinated via a closed I2P channel. Two handles—전창혁 2 (Jeon Chang-hyuk 2) and 최아린 (Choi A-rin)—previously observed running covert operations for the threat actors were active in the channel
2
. Kumar Ritesh, CEO of CYFIRMA, noted the scanning was "deliberately low-volume and paced" with nodes rotated and abandoned at the first sign of blocking.Related Stories
Cloudflare detected millions of exploitation attempts after public disclosure of the PHP vulnerabilities, with tens of thousands occurring within 45 minutes of the announcement
3
. The FBI issued a public service announcement Friday stating "security updates are available and should be applied immediately"3
. The NSA, FBI, and CISA jointly urged organizations Wednesday to patch immediately and take systems offline if they could not.Government websites in Britain, Finland, and Japan experienced outages or limited service during patching work. Broome County, New York, took its website offline Friday "out of an abundance of caution"
3
. PHP released fixes on October 21 for versions 8.2 through 8.5, addressing vulnerabilities that affect as much as 70% of websites—a period some cybersecurity professionals have called "Black October."The PHP2Shell vulnerabilities were initially discovered using artificial intelligence. In August, Sagi Tzadik, a researcher at Google-owned Wiz, asked an AI agent to adapt a WordPress exploit for Drupal. While the agent failed at that task, it uncovered a PHP flaw that had gone unnoticed for more than a decade
3
. Separately, Calif working with OpenAI found a related bug, with about 15 people collaborating to develop a fix.
Source: Hacker News
The PHP Foundation acknowledged the issue had been known since at least 2024 and described the delay in addressing it as "a structural failure"
3
. Information about the flaws spread two days before the planned patch date, prompting PHP maintainers to release the fix one day early. Hours after the public announcement, Calif mistakenly posted a video demonstrating its exploit, further accelerating exploit attempts.Former CISA Director Chris Krebs urged smaller companies to scan for vulnerabilities "constantly" and apply patches as quickly as possible, calling this "a significant week or two in cybersecurity"
3
. Security experts recommend enabling automatic updates for devices, using multi-factor authentication, and maintaining different passwords for every website as fundamental protective measures against ongoing exploit attempts targeting these critical PHP vulnerabilities.Summarized by
Navi
[2]
22 Oct 2026•Technology

23 Oct 2026•Policy and Regulation

27 Oct 2026•Technology
