tl;dr
- PHP fixed two pre-auth remote code execution bugs on 21 October 2026 after the disclosure leaked a day early; Amazon logged an attempt four days before.
- One documented intrusion chained them with a Linux root bug to take 24 firewalls and a server behind them, blamed on suspected Chinese operators.
- The North Korea link is CYFIRMA’s moderate-confidence read of a 216-host target list; no victim is reported.
Four days before the patch
On 17 October 2026, four days before PHP shipped the fix, a honeypot run by Amazon’s threat intelligence team (MadPot) logged a request that tried to trigger one of the bugs. The request came from a commercial anonymisation network, and Amazon could not say whether it worked.
The bugs did not stay secret until the planned release either. PHP’s security team had arranged a coordinated disclosure, but the embargo broke: the security firm Chainguard published ahead of schedule, a researcher at Calif, another security firm, posted a proof-of-concept video by mistake, and the release was pulled forward a day. On 21 October, PHP shipped fixes for two unauthenticated remote code execution flaws, quickly nicknamed PHP2Shell, and published the full write-ups.
Both bugs live in the PHP runtime itself, not in anyone’s application code. Roughly 70% of websites run PHP, according to W3Techs figures cited by BleepingComputer. A web server handles a request in one of two ways: with mod_php, where the PHP interpreter is loaded into the Apache worker process, or with PHP-FPM, a pool of separate PHP worker processes that the web server hands requests to. Either way a pool of long-lived workers serves request after request.
Both flaws are use-after-free bugs, where code keeps using a chunk of memory after it has been freed and possibly handed to something else. They are reached when a request runs out of memory and PHP aborts it mid-flight: the runtime jumps out of the half-finished request with a longjmp(), and the clean-up that follows walks over state that is no longer valid. CVE-2026-109618 corrupts memory during that bail-out and was demonstrated against Apache mod_php before any application code ran. CVE-2026-109619 leaves a piece of global state pointing into the freed request, so the next request on the same worker trips over it; it was shown to work on both mod_php and PHP-FPM. Neither has a complete configuration workaround. Updating is the only fix.
The way in is a PHP feature called session upload progress, which runs while a request body is still being parsed, before any login check: exactly the window these flaws need.
128 posts, and they don’t agree
Obstracts collected 128 posts on the flaws between 20 and 29 October. Most repeat each other, and they disagree on the basics. Depending on which you read, the severity is 9.2 or 9.8, and the affected versions are PHP 8.2 to 8.5, PHP 5.4 and up, or every 7.x and 8.x release. Some led with a confirmed breach that was later retracted.
A smaller set saw something first-hand:
| Source | What it saw | What it lacks |
|---|---|---|
| PHP security team (21 Oct) | The two bugs, the mechanism, and the fixed versions | Any field data |
| CISA (21 and 26 Oct) | The emergency alert, then the KEV listing | Any description of victims |
| Amazon (21 Oct) | The 17 October attempt, four days early, from an anonymisation network | Whether it succeeded |
| F5 Labs (22 Oct) | The first mass scanning, its user-agent strings and probe markers | Any confirmed shell |
| SOCRadar (22 Oct) | An attacker staging server and its exploit script | A single success |
| Unit 42 (23 to 29 Oct) | Exposure counts and hunting logic; a breach claim it later pulled | A confirmed breach |
| Wiz (21 and 27 Oct) | Internet-wide exposure trends and attempted exploitation | A success in its own telemetry, as of 27 Oct |
| Adel Karimi, via BleepingComputer (26 Oct) | A confirmed shell on a deliberately favourable mod_php honeypot | Any sign it generalises |
| CYFIRMA, via Security Affairs (26 Oct) | Underground staging and a suspected North Korea link | Any victim; its own advisory (this rests on Security Affairs) |
| OffSec (28 Oct) | The request-lifecycle anatomy of both bugs and the signals to hunt for | Field telemetry |
| Dream, via The Register (29 Oct) | A custom chain that rooted 24 firewalls and an internal management server behind them | Published indicators; its own write-up (this rests on The Register and LetsDataScience) |
They overlap on the bugs and on the warning to hunt for PHP workers spawning shells. They split on everything that followed: whether anything had actually been exploited, by whom, and how seriously.
From probes to working shells
Mass scanning started the morning the patch landed. F5 logged the first attempts at 08:36 UTC on 21 October, almost all of them aimed at CVE-2026-109619, carrying obvious user-agent strings such as php2shell-verify/1.0 and php2shell-scanner/1.0 and a fixed probe marker in the request. SOCRadar traced a staging server at 206.188.197[.]105 serving an aslr_exploit.py written for PHP 8.4.26 under mod_php. It recorded no success, and tied the server to nobody.
For the first few days that was the pattern: a great deal of noise, nothing landing. The exception was a honeypot. On 24 October, Adel Karimi, of Perplexity, recorded a working shell against Apache mod_php on PHP 8.4.26, on what he described as an intentionally favourable single-worker profile. Marcus Hutchins of Expel called it by far the highest-quality confirmation so far. According to GovInfoSecurity, six of Karimi’s honeypots recorded 35 successful exploits from two addresses. Two days later, on 26 October, CISA added both CVEs to its Known Exploited Vulnerabilities list and, according to The Record, ordered civilian federal agencies to patch by that Thursday. CISA’s bar for that list includes attempted exploitation, and it said nothing about any victim; our read is that the honeypot shells were the likely trigger.
On 27 October Wiz said that none of the attempts in its own telemetry had succeeded, and that exposed instances had dropped by about two thirds as operators patched or pulled servers offline. It named the two addresses behind Karimi’s shells, 15.204.59[.]70 and 162.35.99[.]94. It also found that some tools branded as PHP2Shell exploits were doing something else entirely: poisoning a session file and pulling it in through a local file include, a flaw that tricks an application into loading a file of the attacker’s choosing, rather than triggering either CVE.
Breaches that weren’t
Along the way, two confirmed breaches were announced and then withdrawn. On 23 October, Unit 42, the threat research team at Palo Alto Networks, whose firewalls appear later in this story, described agentic-AI post-exploitation activity tied to the flaws, with command batches wrapped in delimiters such as ==SHELLUSER== and ==PASSWD== for an AI model to read, as The Hacker News quoted it; on 25 October it retracted the link, saying it was unable to confirm that the initial access was associated with the specific PHP2Shell CVEs. Sophos made, and withdrew, a similar claim the same week, The Record reported.
Both still belong in the record. The next analyst who hears that PHP2Shell was breaching organisations within 48 hours should be able to find the claim, who made it and why it was pulled.
Who is blamed, and on what evidence
By 29 October, two countries were being blamed for PHP2Shell, and they point in different directions.
The one documented intrusion is attributed to suspected Chinese operators. Dream, the Israeli AI cybersecurity startup that reported it, describes the attackers as suspected Chinese, according to The Register, which lists the targeted countries as the US, Japan, Taiwan and South Korea. The news site LetsDataScience calls the China-nexus link an assessment, not a confirmed identification, and reports that Unit 42 tracks the activity as CL-UNK-1280 and attributes it to a Chinese-speaking attacker in the UTC+8 time zone. That is second-hand: Unit 42’s own brief doesn’t name the cluster, though it says it independently verified the attacker’s tooling. Treat the attribution as Dream’s, relayed.
The louder name is North Korea, and it rests on far less. CYFIRMA, a threat intelligence vendor that monitored a private channel from 22 October, described a campaign staging to exploit the flaws: 216 target hosts, 21 scanning nodes and a local AI model ranking exploits. It linked the activity to Lazarus-family groups at moderate confidence while noting that a Korean-speaking criminal group would fit the evidence just as well, named South Korea among more than ten countries on its target list, and reported no victims. This is a target list and a staging operation; nobody has reported an intrusion. Two things that actually happened are attributed to nobody: the 17 October attempt on Amazon’s honeypot, and the shells on Karimi’s honeypots.
Root in under fifteen minutes
Dream’s intrusion is the only end-to-end attack on record, and it did not touch a normal website. The operators built a custom exploit for the management interface of Palo Alto Networks PAN-OS firewalls, which run PHP 8.0.30. They found roughly 19,000 internet-facing firewalls with the internet-scanning service FOFA (per The Register), narrowed them to a target list of 240 (per LetsDataScience), and rooted 24 firewalls and an internal management server behind them, 25 servers in all. Two of the firewalls were on a US government-affiliated network: the first fell on 23 October but crashed; the second, on 24 October, gave them a foothold. Start to root took under fifteen minutes.
The steps below follow that reporting of Dream; the step order is theirs, not confirmed against Dream’s own write-up. If Attack Flows are new to you, Beyond the ATT&CK Matrix: How to Build Dynamic Attack Flows with STIX explains how to read one, and Understanding STIX 2.1 Objects covers the objects in the graphs.
- They find exposed management interfaces with FOFA (Scan Databases, T1596.005).
- They fire the custom PHP2Shell exploit at the management interface (Exploit Public-Facing Application, T1190). ATT&CK has no technique for how the exploit itself works, so the graph spells it out at this step. As The Register describes it, crafted multipart POST requests corrupt PHP’s memory into a predictable layout; a second request forges the length of a stale session ID, causing an out-of-bounds read that exposes the base address of the PHP shared library,
libphp.so; and exhausting memory again leaves a stale internal reference that the final stage forges to run commands. - The exploit runs commands as the unprivileged
nobodyuser the web service runs as (Unix Shell, T1059.004). - They chain a Linux flaw, Fragnesia (CVE-2026-46300), to escalate from
nobodyto root (Exploitation for Privilege Escalation, T1068). - They plant a root-privileged webshell in a public directory (Web Shell, T1505.003).
- They open a reverse SSH tunnel from a rooted firewall to reach an internal management server (Protocol Tunneling, T1572).
- They decrypt the stored password of the LDAP service account, the account the firewall uses to query the directory, using the firewall’s master key (Credentials In Files, T1552.001).
- They enumerate the directory’s accounts and groups (Domain Account and Domain Groups discovery, T1087.002 and T1069.002).
- They spray passwords against those accounts (Password Spraying, T1110.003).
- They try to create backdoor domain accounts (Create Account, T1136.002); the outcome is unknown.
- They try to steal the SAML signing key, which would let them forge single sign-on logins (Private Keys, T1552.004); the outcome is unknown.
The firewall was the way in, not the target. Steps 7 to 11 are an attempt to turn one rooted appliance into access across the network behind it. CISA treated the chain seriously enough to add Fragnesia to the KEV list on 28 October, calling out that it can be chained with the two PHP flaws to run commands as root.
What nobody has confirmed
Several things sit unresolved on 29 October. Nobody has explained how the 17 October attacker had a working exploit four days before the bugs were public; Amazon’s full request, or the earliest sample of a working exploit, would show whether that attacker had the leaked details or found the bugs independently. CVE-2026-109618 is demonstrated on mod_php but only called “susceptible” on PHP-FPM, with no full exploit shown. Whether CYFIRMA’s staging ever became an intrusion is unknown. AI’s role is suggestive but unproven: Dream found Claude Code and an “aiscan” tool wired to DeepSeek and Kimi models on the attacker side, but no direct evidence that AI wrote the exploit. And a third PHP flaw, CVE-2026-115560, which BleepingComputer reported needs an unusual mix of configuration and application code, was still unpatched; the advice was to turn session upload progress off.
What defenders can hunt for
The indicators published so far are leads, not blocklist entries. Amazon released six addresses, one seen on 17 October and five on 21 October, and told readers to investigate rather than block them, because the list includes shared anonymisation infrastructure. Wiz published nine, including 31.56.197[.]72 as a payload server and the two addresses that landed shells on Karimi’s honeypots. Alongside the addresses, the branded scanning left fingerprints worth grepping for: user-agents like php2shell-scanner/1.0, F5’s probe value p2sprobec982a06c in a PHP_SESSION_UPLOAD_PROGRESS field, and the p2s-rider marker in the same field, which Defused Cyber, a threat intelligence company that runs honeypots, saw from 204.76.203[.]7 on its honeypots for GlobalProtect, Palo Alto Networks’ VPN gateway, as reported by The Hacker News. The graph shows the addresses only; the user-agents and markers are in the reference section.
The behaviour is steadier than any address. CISA, the FBI and Unit 42 say to watch for a PHP worker spawning a shell or downloader; OffSec stresses the shell, and Amazon advises checking for unexpected process execution. A successful exploit ends with the nobody user running sh, curl or wget, which a normal PHP request almost never does. A Sigma rule is a YAML file that describes suspicious activity in a log, which detection tools convert into queries for a specific SIEM. If you are new to Sigma, Writing Effective Sigma Detection Rules: A Guide for Novice Detection Engineers explains each field. This one watches Linux process creation for a web or PHP worker launching a shell, a downloader or the PHP command-line binary:
title: Web or PHP Worker Spawning a Shell or Downloader on Linux
id: 24c7ce67-c368-5a86-abcc-13056acd8b34
status: experimental
description: Detects a PHP-serving process (Apache with mod_php, or a PHP-FPM worker) spawning a shell or a download utility. In the PHP2Shell intrusions of October 2026, a successful exploit of CVE-2026-109618 or CVE-2026-109619 ran commands as the unprivileged web user, which a normal PHP request does not do.
references:
- https://www.cisa.gov/news-events/alerts/2026/10/21/critical-vulnerabilities-php-require-immediate-patching
- https://www.ic3.gov/PSA/2026/PSA261023
- https://www.offsec.com/blog/php2shell-anatomy-of-two-php-lifecycle-bugs/
author: dogesec
date: 2026-10-29
tags:
- attack.execution
- attack.t1059.004
logsource:
category: process_creation
product: linux
detection:
parent_web:
ParentImage|endswith:
- '/apache2'
- '/httpd'
parent_fpm:
ParentImage|contains: 'php-fpm'
child_shell:
Image|endswith:
- '/sh'
- '/bash'
- '/dash'
- '/wget'
- '/curl'
child_php:
Image|re: '/php[0-9.]*$'
condition: (parent_web or parent_fpm) and (child_shell or child_php)
falsepositives:
- CMS plugins or cron wrappers that legitimately shell out, such as backup scripts, ImageMagick image processing, or git deploy hooks
- Developer tooling running under the web user on a non-production host
level: high
The rule catches step 3, the moment the exploit runs a command. It cannot see the exploit itself: as OffSec puts it, a successful attempt can return a clean HTTP 200 and leave the worker standing, so a quiet log proves nothing. Two things make a weaker but useful second signal in the error log: the message Allowed memory size ... exhausted ... in Unknown on line 0, and Apache AH00052 segfault notices, both of which the failed attempts tend to leave. And on a PAN-OS firewall there is often no endpoint agent watching process creation at all, so this rule may have nothing to run on.
The whole case in one graph
Put together, this is the whole case: the first-hand reports, the two names and the evidence behind each, the firewall chain in order, with the exploit mechanics at step 2, the indicator leads and the detection, and the questions still open on 29 October.
A hunter gets one thing to deploy today: a web worker spawning a shell. An analyst gets the state of the case on 29 October, with every claim traceable to who made it.
In summary
PHP2Shell produced 128 posts in ten days; most of them repeated each other, and some led with a breach that was later pulled. Stripped back, the facts are smaller and sharper: two runtime bugs behind most of the web, made public a day early, and probed four days before the patch, which produced huge scanning and almost no successes. The one real intrusion did not touch a website at all; it chained the flaws with a Linux root bug to take over firewall management interfaces and reach the network behind them. Patch, and until you have, hunt on your PHP servers for what every confirmed exploit so far has ended with: a PHP worker running a shell.
Reference: IoCs and ATT&CK techniques
IoCs
Every value below is also in one downloadable bundle.
| Type | Value | What it is | Linked to | Confidence | STIX ID |
|---|---|---|---|---|---|
| IPv4 | 206.188.197[.]105 |
SOCRadar: attacker staging server (ports 9999, 8899, 8888) serving aslr_exploit.py for PHP 8.4.26 under mod_php. SOCRadar recorded no success and tied it to nobody. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--39dfb6ee-b5cb-5fe5-b2ac-3bfe5f5df39f |
| IPv4 | 103.216.220[.]92 |
Amazon (MadPot): the 17 October 2026 exploit attempt, four days before the fix, from a commercial anonymisation network; Amazon advises investigating rather than blocking its six addresses. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--a031c6df-070e-574a-9dcc-9065cbddfdb2 |
| IPv4 | 31.223.184[.]166 |
Amazon (MadPot): seen in PHP2Shell exploitation on 21 October 2026; one of six addresses Amazon says include shared anonymisation infrastructure, so it advises investigating rather than blocking them. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--f48f7382-6ea9-57e6-8b79-07e08a143e5b |
| IPv4 | 37.19.200[.]5 |
Amazon (MadPot): seen in PHP2Shell exploitation on 21 October 2026; one of six addresses Amazon says include shared anonymisation infrastructure, so it advises investigating rather than blocking them. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--3cf5eab6-9e78-5022-9e37-43fa670dedb8 |
| IPv4 | 134.209.159[.]64 |
Amazon (MadPot): seen in PHP2Shell exploitation on 21 October 2026; one of six addresses Amazon says include shared anonymisation infrastructure, so it advises investigating rather than blocking them. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--407b2ec2-09ee-5662-b79b-83b5c93cd8be |
| IPv4 | 103.114.163[.]188 |
Amazon (MadPot): seen in PHP2Shell exploitation on 21 October 2026; one of six addresses Amazon says include shared anonymisation infrastructure, so it advises investigating rather than blocking them. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--5c4beb38-b4f2-501a-aab5-5e8f86abb538 |
| IPv4 | 95.217.98[.]37 |
Amazon (MadPot): seen in PHP2Shell exploitation on 21 October 2026; one of six addresses Amazon says include shared anonymisation infrastructure, so it advises investigating rather than blocking them. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--a31cb5cc-a4be-5c45-baeb-2f8d65610744 |
| IPv4 | 31.56.197[.]72 |
Wiz: C2 (payload server). | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--ec279506-970a-5912-88ea-5658014fd5ac |
| IPv4 | 162.35.99[.]94 |
Wiz: exploitation attempt; one of the two addresses Wiz says successfully exploited Adel Karimi’s honeypots. The captured commands fetched payloads from it with wget, curl or the PHP CLI (The Hacker News). | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--38ff7f3a-a272-5fd8-9553-ee9f32ee7edd |
| IPv4 | 15.204.59[.]70 |
Wiz: exploitation attempt; one of the two addresses Wiz says successfully exploited Adel Karimi’s honeypots. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--38a27e86-9dcf-5f80-b6b9-33d189fa7c3e |
| IPv4 | 85.11.187[.]35 |
Wiz: exploitation attempt. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--aaa2547e-bfda-516b-b962-45b237a3a0bd |
| IPv4 | 172.206.2[.]199 |
Wiz: scanner and exploitation attempt. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--164c0704-6f4f-5952-b406-825462b59b46 |
| IPv4 | 3.37.227[.]134 |
Wiz: exploitation attempt. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--9ec3fbf3-150f-5b6b-a9e0-729d732541c5 |
| IPv4 | 103.151.172[.]95 |
Wiz: exploitation attempt. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--822d625b-8cf2-562f-900b-2a9eb9f5639e |
| IPv4 | 178.249.214[.]6 |
Wiz: exploitation attempt (VPN). | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--c9ecc190-cd3c-5b6c-a6ed-d16b284ba94a |
| IPv4 | 135.136.19[.]197 |
Wiz: scanner and exploitation attempt (VPN). | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--2a988c9a-e173-53d0-98aa-533a25593ec5 |
| IPv4 | 204.76.203[.]7 |
Defused Cyber (via The Hacker News): source of PHP2Shell probes carrying the p2s-rider marker against its Palo Alto Networks GlobalProtect honeypots; The Hacker News calls it a known mass-exploit address. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--9114ada1-faeb-5087-8463-6c04cce8a8db |
| User-agent | php2shell-scanner/1.0 |
F5 Labs: user-agent seen in PHP2Shell scanning from 21 October 2026. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--27479dbd-d7f2-5b37-a8df-26a860b7f40d |
| User-agent | php2shell-verify/1.0 |
F5 Labs: user-agent seen in PHP2Shell scanning from 21 October 2026. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--5269db7d-1499-531c-9bab-b30a9687f5b4 |
| User-agent | php2shell |
F5 Labs: user-agent seen in PHP2Shell scanning from 21 October 2026. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--59a2638e-ced9-5c31-a77e-e0d222db278e |
| User-agent | php-upload-progress-scan/1.0 (internal, authorized-use-only) |
F5 Labs: user-agent seen in PHP2Shell scanning from 21 October 2026. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--d835f1b4-722c-54ba-a30d-83c3fbef9893 |
| String | p2sprobec982a06c |
F5 Labs: fixed probe value sent in the PHP_SESSION_UPLOAD_PROGRESS field of PHP2Shell scanning requests. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--f40b0101-786e-551c-b9fa-64d8b00ef60f |
| String | p2s-rider |
Defused Cyber (via The Hacker News): marker inside a PHP_SESSION_UPLOAD_PROGRESS field in PHP2Shell probes against its GlobalProtect honeypots, from 204.76.203[.]7. | PHP2Shell scanning, exploitation attempts and staging (unattributed) | Low | indicator--b9e579f0-6d27-54b2-abd6-fa4ef4355266 |
Techniques (firewall intrusion Attack Flow)
| Step | Action | Technique | Tactic | STIX ID |
|---|---|---|---|---|
| 1 | Exposed management interfaces found with FOFA | T1596.005 | Reconnaissance (TA0043) | attack-action--cfb81474-0356-5ade-83fd-d177363006b4 |
| 2 | Custom PHP2Shell exploit against the management interface | T1190 | Initial Access (TA0001) | attack-action--ee0fc077-8b6b-5ed6-9074-b2ff49b2b86f |
| 2a | Heap groom and libphp base-address leak via a stale session ID | None (procedure) | Initial Access (TA0001) | procedure--5d63d3fc-8a90-55be-b4a1-769562517629 |
| 3 | Commands run as the nobody web user |
T1059.004 | Execution (TA0002) | attack-action--16448c16-f59c-56c4-9f7e-81371ae95b89 |
| 4 | Fragnesia (CVE-2026-46300) chained to root | T1068 | Privilege Escalation (TA0004) | attack-action--87ee2b2e-c472-5371-8246-7dd69b85ab43 |
| 5 | Root webshell planted in a public directory | T1505.003 | Persistence (TA0003) | attack-action--9438bef7-cda6-58cf-bc5c-70d4c21fc478 |
| 6 | Reverse SSH tunnel to an internal management server | T1572 | Command and Control (TA0011) | attack-action--218a49ab-0a24-5a40-a4f9-6240aabea57c |
| 7 | LDAP service-account password decrypted with the master key | T1552.001 | Credential Access (TA0006) | attack-action--01b5d09e-9618-5629-9648-7f765d619822 |
| 8 | Directory accounts and groups enumerated | T1087.002, T1069.002 | Discovery (TA0007) | attack-action--e7f88597-f0f1-5578-9f0c-7444c8f382e9 |
| 9 | Password spraying against those accounts | T1110.003 | Credential Access (TA0006) | attack-action--7023734b-20e7-5046-b38e-6f149f4fdda3 |
| 10 | Attempted backdoor domain accounts (outcome unknown) | T1136.002 | Persistence (TA0003) | attack-action--b0959928-e382-5181-82e6-7fd0070b0a43 |
| 11 | Attempted SAML signing-key theft (outcome unknown) | T1552.004 | Credential Access (TA0006) | attack-action--c8e70ca2-282d-504b-b9b2-fb3381401869 |
The Sigma rule is Sigma ID 24c7ce67-c368-5a86-abcc-13056acd8b34 (STIX indicator--24c7ce67-c368-5a86-abcc-13056acd8b34), and the firewall campaign is campaign--42fb7a2a-8115-582b-9f19-ba9193a970e0.
Get new posts by email
New threat research and tutorials from dogesec, plus the weekly Obstracts threat brief. No spam, unsubscribe any time.
