🔍 Read the full analysis: Inside The PHP2Shell Chain: How 25 Palo Alto Servers Were Taken Over on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Dream says an attacker used PHP2Shell flaws and a previously disclosed Linux kernel vulnerability to gain root access on 24 Palo Alto Networks firewalls, then compromise an internal firewall-management server in a U.S. government-affiliated network. Palo Alto Networks says Unit 42 independently verified the tools and results. Dream found AI tools in the attacker’s environment but says it has no direct evidence they created the exploit.
An attacker used two PHP2Shell vulnerabilities and a known Linux kernel flaw to gain root access on 24 Palo Alto Networks firewalls, then reached an internal management server in a U.S. government-affiliated network, according to Israeli security firm Dream. Palo Alto Networks said its Unit 42 researchers independently verified the attacker’s tools and results; Dream’s account describes 25 compromised servers in total.
Dream said the attacker ran an automated campaign against 240 internet-reachable PAN-OS firewalls on the evening of October 23. The company reported that 24 were compromised with root access. The targets were located in the United States, Japan, Taiwan and South Korea. Dream said 22 of the rooted firewalls were left with hidden web shells, while two belonged to the government-affiliated organization later targeted from within.
According to Dream’s timeline, the attacker used a firewall as a route into that organization’s network on October 24, compromised its internal server for managing firewalls, and then targeted Active Directory, the system used to manage identities and access. Dream said the attacker established a tunnel and persistence that survived reboots. The source account does not state whether the Active Directory attempt succeeded.
The reported chain combined PHP2Shell vulnerabilities CVE-2026-109618 and CVE-2026-109619 with Fragnesia, a Linux kernel privilege-escalation flaw identified as CVE-2026-46300. Dream said PHP2Shell enabled command execution under a restricted account, while Fragnesia was used to obtain root privileges. The company said the full chain to a persistent back door took less than 15 minutes. That timing is Dream’s account, not a separately established measure for every compromised device.
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
From Exposed Firewalls to Internal Access
The incident shows how a device intended to enforce network boundaries can become an entry point when its management interface is exposed to the internet and its software remains vulnerable. In the government-affiliated network described by Dream, compromising perimeter firewalls provided a path to an internal server that managed other firewalls. The subsequent targeting of Active Directory raises the possibility of wider access, though Dream’s material does not confirm that the identity system was compromised.
The case also highlights the short window between vulnerability fixes becoming available upstream and vendor-specific fixes reaching affected products. PHP released corrected versions on October 21, Dream said, while fixes for the affected PAN-OS branches arrived later. For operators, applying a patch is only one part of reducing exposure: restricting management access can limit risk while a vendor fix is pending. The reported compromise is evidence of an attack against specific systems, not proof that every exposed firewall was affected.
Dream found Claude Code installed on attacker infrastructure and DeepSeek and Kimi configured in an offensive-security platform called aiscan. That indicates AI tools were present in the environment, but not what role they played. The distinction matters because attributing an exploit to AI requires evidence that the tools generated or materially shaped it, rather than simply being available to the operator.
Products Worth Considering

Dell UltraSharp U4025QW 40-Inch Curved 5K2K WUHD LED Monitor

Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER), QHD

Samsung 32-Inch ViewFinity S7 4K UHD IPS Monitor
As an affiliate, we earn on qualifying purchases.
Patches, Vulnerabilities and Attack Timeline
Fragnesia was publicly disclosed in May 2026, according to the supplied account, and a public exploit was available. Dream said the attacker set up aiscan with DeepSeek and Kimi on October 8. Before the attacks, researchers reportedly identified about 19,000 PAN-OS firewalls with internet-facing management pages using FOFA, an internet-connected-device search engine. A custom scanner obtained version information from about 5,000; Dream said the attacker narrowed that pool to 240 targets based on country and software version.
On October 21, the PHP project released updated versions of PHP and Palo Alto Networks published an advisory, but fixed PAN-OS builds were still forthcoming. Dream says the company’s PAN-OS appliance software used its own copy of PHP, version 8.0.30. The supplied report places the attack batch between 18:14 and 21:55 UTC on October 23. A government-affiliated firewall was among the first compromised; Dream said it crashed shortly afterward, possibly following about 900 megabytes of attack traffic. The cause is presented as likely, not confirmed.
Dream published its findings on October 26 and updated them on October 29. The source material says the U.S. Cybersecurity and Infrastructure Security Agency added both PHP2Shell vulnerabilities to its Known Exploited Vulnerabilities catalog on October 26 and set a three-day deadline for federal agencies. Palo Alto Networks’ fixes for the affected 12.1 and 12.2 branches were released between October 27 and 29, according to the account.
““no direct evidence” that AI tools wrote the exploit or scripts”
— Dream
AI Role and Intrusion Scope
Dream says it found AI tools on the attacker’s servers, but it has no direct evidence that those tools wrote the exploit or attack scripts. Their presence does not establish that AI developed the vulnerability chain, selected targets, or operated the campaign. Dream also said the activity did not appear fully autonomous.
The supplied account does not identify the affected organizations beyond describing one as U.S. government-affiliated, nor does it say whether data was taken, whether the Active Directory attack succeeded, or whether all hidden web shells and persistence mechanisms have been removed. The reported 24 firewall compromises and one internal server compromise are findings attributed to Dream, with Palo Alto Networks saying Unit 42 verified the tools and results. The precise effects on each organization remain unclear.
It is also not clear from the material how many of the approximately 19,000 exposed devices were vulnerable to the relevant version-specific chain, or how many have since been patched or taken off the public internet. The report’s estimate that the first government-affiliated firewall crashed because of attack traffic is also not confirmed as the cause.
Patch and Restrict Management Access
Organizations using affected PAN-OS branches should check Palo Alto Networks’ current advisories and apply the fixed versions relevant to their deployments. Dream’s account says fixed builds for branches 12.1 and 12.2 shipped between October 27 and 29, 2026; administrators should verify their exact software version and vendor guidance rather than assume a device is protected based on the branch name alone.
While checking patch status, network operators should review whether firewall management interfaces are reachable from the public internet and limit access to trusted, controlled networks. Organizations that may have been exposed should investigate for web shells, unauthorized administrator changes, unexpected tunnels and persistence, and review authentication activity around the reported dates. The source material does not describe the outcomes of remediation or any further public findings, so updates from Dream, Palo Alto Networks and relevant government agencies may clarify the extent of the incident.
Key Questions
What does PHP2Shell refer to in this report?
PHP2Shell is the name used for two vulnerabilities, CVE-2026-109618 and CVE-2026-109619. Dream says the attacker used them to run commands on affected firewalls before chaining in a Linux kernel flaw to obtain root access.
How many servers did Dream say were compromised?
Dream reported 24 PAN-OS firewalls and one internal firewall-management server, for 25 servers overall. Palo Alto Networks said Unit 42 independently verified the attacker’s tools and results.
Did AI create or run the attack?
That is not established. Dream found Claude Code and configured DeepSeek and Kimi models on attacker infrastructure, but said it had no direct evidence that AI wrote the exploit or scripts and that the operation did not appear fully autonomous.
What should administrators do?
Administrators should review Palo Alto Networks’ advisories, install the fixed PAN-OS versions applicable to their systems, and restrict firewall management access from the public internet. Potentially exposed organizations should also investigate for unauthorized access and persistence.
Source: ThorstenMeyerAI.com
