🔍 Read the full analysis: When “AI-Assisted” Meets PHP2Shell: Inside A Two-Day Server Takeover on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Dream reported that an attacker used PHP2Shell and a previously disclosed Linux kernel flaw to gain root access on 24 Palo Alto Networks firewalls, then reached a 25th server inside a US government-affiliated network. Palo Alto Networks says Unit 42 independently verified the attacker’s tools and results. Dream found AI tools on the attacker’s servers but said it has no direct evidence they wrote the exploit or scripts.
An attacker gained root access to 24 Palo Alto Networks firewalls and used one foothold to reach a 25th server inside a US government-affiliated network, according to a report by Israeli security company Dream. The October 23-24 attacks followed the release of fixes for two PHP flaws known as PHP2Shell; Palo Alto Networks says its Unit 42 researchers independently verified the attacker’s tools and results.
Dream said the attacker selected 240 internet-reachable firewalls and rooted 24 of them. The reported targets were in the United States, Japan, Taiwan and South Korea. On 22 compromised firewalls, Dream found a hidden web shell left in place. The report says two compromised firewalls belonged to the government-affiliated organization; the attacker used them to reach an internal server that managed the organization’s firewalls.
The attack combined PHP2Shell vulnerabilities CVE-2026-109618 and CVE-2026-109619 with Fragnesia, a Linux kernel privilege-escalation flaw, CVE-2026-46300. Dream’s account says PHP2Shell enabled command execution as a low-privilege user, while Fragnesia helped the attacker gain administrator-level root access. Dream estimated that the chain, including installation of a persistent back door, could take under 15 minutes.
Dream reported that the attacker reached the internal management server on October 24 and then targeted Active Directory, the organization’s system for managing user accounts and access. Dream’s published account identifies 25 compromised servers in total: 24 firewalls and one internal server. It does not establish the full consequences of the Active Directory activity.
Infographic 01 · PHP2Shell on Palo Alto firewalls
Two days from patch to root.
PHP fixed the bug on 21 October. Fixed firewall software for the attacked branches arrived almost a week later. The attack ran in between.
- 13–23 MayFragnesia goes public, kernel fix followsLinux privilege-escalation bug with a public exploit; CVE-2026-46300.
- 8 OctAttacker sets up an AI platformaiscan configured with DeepSeek and Kimi, per Dream’s timeline.
- 21 OctPHP fixes PHP2ShellPHP 8.5.12, 8.4.27, 8.3.36, 8.2.35. Palo Alto advisory the same day; PAN-OS fixes “upcoming”.
- 23 Oct, evening UTCBatch run: 240 targets, 24 rootedFound via device search engine, filtered to US, Japan, Taiwan, South Korea.
- 24 OctInto a government-affiliated networkTunnel, persistence, internal management server rooted (no. 25), Active Directory attacked.
- 26 OctDream publishes · CISA lists PHP2ShellFederal deadline 29 Oct.
- 27–29 OctFixed PAN-OS builds for 12.2 and 12.1CISA lists Fragnesia on 28 Oct; Unit 42 confirms Dream’s results on 29 Oct.
The fix for PHP existed. The fix for the firewall did not, yet. The open management page was the only door that mattered.Thorsten Meyer · my reading
Sources: Dream Research, 26 Oct 2026 (updated 29 Oct); php.net, 21 Oct 2026; Palo Alto Networks advisory (updated 29 Oct 2026); CISA KEV catalog, version 29 Oct 2026; oss-security, 13 May 2026; Unit 42, 29 Oct 2026. Day count: Thorsten Meyer.
Powered by Thorsten Meyer AI
Exposed Management Pages Became Entry Points
The incident shows how an internet-facing firewall management interface can turn a newly disclosed software flaw into access to an organization’s internal systems. A firewall is meant to control network traffic, but its management service can itself become an entry point when exposed and vulnerable. Here, Dream says the attacker moved from firewall access to an internal management server and then targeted Active Directory.
The timing also matters for administrators. PHP released fixes on October 21, but Dream says PAN-OS uses its own PHP copy, and Palo Alto’s October 21 advisory described fixed PAN-OS builds as forthcoming. According to the source account, fixes for the affected 12.1 and 12.2 branches arrived between October 27 and 29. In the intervening period, Palo Alto advised keeping management access off the public internet. The report underscores the need to restrict management interfaces and apply vendor fixes promptly, while checking systems for signs of compromise.
Dream also found Claude Code installed and DeepSeek and Kimi configured in an offensive-security platform called aiscan. That is evidence of AI tools present in the attacker’s environment, not proof that AI generated the exploit or ran the operation autonomously. The distinction matters: the tools may have supported preparation or speed, but the report does not establish their role in the attack itself.
Popular right now
Shark PowerDetect Corded Stick Vacuum, Self-Cleaning (Eggplant, HZ4002)
Vitamix Propel Series 750 Professional-Grade Blender, 4 Programs, Black
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
As an affiliate, we earn on qualifying purchases.
Patches Followed Earlier Kernel Disclosure
The reported chain joined flaws with different timelines. Fragnesia, the Linux kernel privilege-escalation bug, was publicly disclosed in May 2026 with an exploit, according to the source account. PHP issued fixes for the two PHP2Shell vulnerabilities on October 21. Palo Alto Networks published an advisory that day, but the account says fixes for the affected PAN-OS branches were not available until October 27-29.
Dream’s timeline says its researchers found roughly 19,000 PAN-OS firewalls with management pages exposed to the internet using FOFA, a search engine for internet-connected devices. A custom scanner obtained version information from about 5,000, and filtering by country and software version produced a list of 240 targets. Those figures describe Dream’s reported scanning and selection process; they are not a count of devices confirmed vulnerable or compromised.
Dream said the attacker configured aiscan with DeepSeek and Kimi on October 8. Researchers found attack artifacts on October 25, and Dream published its report the following day, updating it on October 29. CISA added both PHP2Shell vulnerabilities to its Known Exploited Vulnerabilities catalog on October 26 and set a three-day deadline for federal agencies, according to the supplied account.
““no direct evidence” that AI tools wrote the exploit or scripts”
— Dream
AI’s Role and Data Impact Remain Unknown
The available account does not show whether AI tools generated any code used in the intrusion, helped select targets, or otherwise affected the operation. Dream explicitly said it had no direct evidence that the tools wrote the exploit or scripts, and its findings do not establish a fully autonomous attack.
It is also unclear from the supplied material what information, if any, the attacker accessed or extracted, whether the Active Directory effort succeeded, and whether the organization’s systems or users suffered further effects. The report does not provide a full accounting of the affected organizations beyond the stated countries and the government-affiliated network. The number of exposed devices Dream identified should not be read as a confirmed count of vulnerable or compromised systems.
Patch, Restrict, and Check for Persistence
Organizations using affected PAN-OS branches should follow Palo Alto Networks’ advisories, install the applicable fixed builds, and keep management interfaces off the public internet, as the vendor recommended while fixes were pending. Administrators should also review systems for unauthorized accounts, web shells, tunnels or other persistence, and investigate whether firewall credentials or internal management systems were affected.
Further public reporting or vendor and government notices may clarify the extent of the Active Directory activity and any data impact. The supplied source material does not identify a confirmed next technical disclosure or incident-response milestone. For now, the central unresolved issue is not whether AI was present in the attacker’s environment, but what role, if any, it played in an operation that Dream says relied on exposed management services and known software flaws.
Key Questions
What happened in the PHP2Shell incident?
Dream reported that an attacker combined two PHP vulnerabilities with the Linux kernel flaw Fragnesia to gain root access on 24 PAN-OS firewalls. The attacker then used access through two firewalls to reach an internal server at a US government-affiliated organization.
Were AI tools proven to have written the attack?
No. Dream found Claude Code installed and DeepSeek and Kimi configured in an offensive-security platform, but said it had no direct evidence that AI tools wrote the exploit or scripts.
How many servers did Dream say were compromised?
Dream reported 25 servers: 24 firewalls and one internal firewall-management server. It said 22 of the rooted firewalls were left with a hidden web shell.
What should PAN-OS administrators do?
Install the fixed version that applies to their systems, follow Palo Alto Networks’ security guidance, and restrict firewall management access from the public internet. Administrators should also check for signs of unauthorized access or persistence.
What is still unknown about the attack?
The supplied reporting does not establish whether data was taken, whether the Active Directory attack succeeded, or what specific role AI tools played. It also does not provide a complete account of effects across all targeted organizations.
Source: ThorstenMeyerAI.com
