PHP Vulnerabilities (CVE-2026-109618, CVE-2026-109619) Allow RCE

Critical·Exploited·PoC public·CVSS 9.8·EPSS 100%·23 outletsNew PHP Vulnerabilities (CVE-2026-109618, CVE-2026-109619) Allow RCE Two critical PHP vulnerabilities allow unauthenticated remote code execution via memory handling flaws. Do this today Update to PHP 8.5.12, 8.4.27, 8.3.36, or 8.2.35 immediately; consider WAF protection for unpatchable systems. Read the full story ↓
Critical·PoC public·CVSS 9.2·EPSS 33%·8 outletsNew Fortinet PSIRT Details Multiple CVEs Affecting Node.js Multiple high and medium severity CVEs were detailed for various Node.js versions, including RCE flaws. Do this today Review Node.js versions for all services and apply necessary updates addressing CVE-2026-105482, CVE-2026-105483, CVE-2026-106069, and CVE-2026-111497. Read the full story ↓
High·9 outletsNew LiteLLM Supply Chain Attack: Info-Stealer Malware via PyPI Malicious packages for LiteLLM version 1.107.0 were distributed via PyPI and Docker, stealing credentials and secrets. Do this today Update to LiteLLM version 1.106.2 and rotate all exposed secrets. Read the full story ↓

From the editor

Prioritize patching PHP vulnerabilities CVE-2026-109618 and CVE-2026-109619 to prevent unauthenticated RCE. Review Node.js deployments for multiple CVEs detailed by Fortinet PSIRT. Assess dependencies for LiteLLM version 1.107.0 due to reported PyPI supply chain compromise.

Anmol Singh YadavEditor, PwnSpectrum

Actively Exploited 2

PHP Vulnerabilities (CVE-2026-109618, CVE-2026-109619) Allow RCE Critical·Exploited·PoC public·CVSS 9.8·EPSS 100%·23 outlets·Fortinet PSIRTDay 5NewRead Two critical PHP vulnerabilities allow unauthenticated remote code execution via memory handling flaws.
CVE-2026-109618 allows RCE through improper handling of request memory after Out-Of-Memory errors. CVE-2026-109619 enables RCE due to global state referencing freed request memory.

Why it matters. These vulnerabilities allow unauthenticated attackers to execute arbitrary code on systems running vulnerable PHP versions.

Do this today Update to PHP 8.5.12, 8.4.27, 8.3.36, or 8.2.35 immediately; consider WAF protection for unpatchable systems.

Affected
All PHP installations serving web requests

CVE
CVE-2026-109618, CVE-2026-109619, CVE-2026-115560
1. Thu 22 OctPHP Core RCE Vulnerabilities (CVE-2026-109618, CVE-2026-109619)
2. Fri 23 OctPHP Core RCE Vulnerabilities (CVE-2026-109618, CVE-2026-109619)
3. Sat 24 OctPHP Core RCE Vulnerabilities (CVE-2026-109618, CVE-2026-109619)
4. Mon 26 OctPHP Core RCE Vulnerabilities (CVE-2026-109618, CVE-2026-109619)
5. TodayPHP Vulnerabilities (CVE-2026-109618, CVE-2026-109619) Allow RCE
Read on Fortinet PSIRT Post on X LinkedIn

Also covered by CISA, Unit 42
```
{"coverage":1,"evidence":0.9,"freshness":0.016,"importance":1,"novelty":0.6,"quality":0.95,"total":0.832}
```
CERT-UA warns of Dahua camera exploitation via Easy4IP feature Critical·Exploited·Cip via The RecordNewRead Hackers are exploiting Dahua cameras using the Easy4IP cloud service and default credentials.
CERT-UA identified instances of unauthorized access to Dahua internet-connected cameras and video recorders. Attackers are exploiting the Easy4IP feature, which connects devices via serial numbers to the manufacturer's cloud service. Access can be gained by guessing weak passwords or exploiting flaws in outdated firmware.

Why it matters. Compromised cameras can allow attackers to remotely control devices and maintain access via malicious accounts even after password changes or factory resets.

Do this today Update firmware on Dahua devices to address vulnerabilities CVE-2021-33044 and CVE-2021-33045.

Affected
Dahua internet-connected cameras and video recorders

Read the original · Cip Post on X LinkedIn

Summarised from The Record.
```
{"coverage":0.387,"evidence":0.9,"freshness":0.744,"importance":0.75,"novelty":1,"quality":0.95,"total":0.777}
```

This version contained compiled malware which steals credentials and loads remote-code execution tools

the maintainer, via BleepingComputer

Vulnerabilities & Patches 4

Fortinet PSIRT Details Multiple CVEs Affecting Node.js Critical·PoC public·CVSS 9.2·EPSS 33%·8 outlets·Fortinet PSIRTDay 4NewRead Multiple high and medium severity CVEs were detailed for various Node.js versions, including RCE flaws.
Fortinet PSIRT detailed multiple CVEs affecting Node.js. One vulnerability (CVE-2026-111497) in the Node.js Brotli decoder allows for remote code execution via an out-of-bounds write.

Why it matters. High severity vulnerabilities exist across multiple Node.js versions, providing potential remote code execution pathways for attackers.

Do this today Review Node.js versions for all services and apply necessary updates addressing CVE-2026-105482, CVE-2026-105483, CVE-2026-106069, and CVE-2026-111497.

Affected
Node.js 26.x, 24.x, 22.x

CVE
CVE-2026-105481, CVE-2026-105482, CVE-2026-105483, +7 more
1. Sat 24 OctNode.js RCE via Crafted Brotli Data Demonstrated by Researchers
2. Sun 25 OctNode2Shell: Brotli Decoder Flaw Enables RCE in Node.js and Other Software
3. Mon 26 OctBrotli Decoder Flaw (CVE-2026-111497) Enables RCE Across Multiple Technologies
4. TodayFortinet PSIRT Details Multiple CVEs Affecting Node.js
Read on Fortinet PSIRT Post on X LinkedIn

Also covered by BleepingComputer
```
{"coverage":1,"evidence":0.7,"freshness":0.084,"importance":0.75,"novelty":0.6,"quality":0.95,"total":0.713}
```
Microsoft Authenticator Flaw Allows Token Theft via Android Apps High·CVSS 8.6·Enclave via GBHackersNewRead A flaw in Microsoft Authenticator allows unprivileged Android apps to steal authentication tokens.
CVE-2026-80097 permits a malicious Android application, without requiring permissions, to impersonate trusted apps. The vulnerability stems from an exported activity that trusts identity details supplied by the requesting application.

Why it matters. Successful exploitation allows an attacker to steal plaintext access, refresh, and ID tokens, enabling unauthorized actions against linked work accounts, such as accessing Microsoft Graph resources.

Do this today Ensure all deployed Android devices run Microsoft Authenticator version 6.2606.4246 or later.

Affected
Microsoft Authenticator for Android versions earlier than 6.2606.4246

CVE
CVE-2026-80097

Read the original · Enclave Post on X LinkedIn

Summarised from GBHackers.
```
{"coverage":0.387,"evidence":0.4,"freshness":0.61,"importance":0.75,"novelty":1,"quality":0.65,"total":0.609}
```
Fireblocks mpc-lib: Hardened Derivation Fails Ancestor Isolation (CVE-2026-90975) Medium·CERT/CCNewRead A cryptographic flaw in mpc-lib's hardened derivation allows compromised shares to recover ancestor shares.
The mpc-lib library has a fail-open vulnerability in its BIP32 hardened derivation when a parent key is missing. This failure substitutes the parent contribution with a null buffer, making the derivation delta computable.

Why it matters. This breaks ancestor isolation, meaning a compromise of a lower-level signer share can allow recovery of higher-level shares up the key hierarchy. This degrades a defense-in-depth layer silently.

Do this today Treat the chain code as secret and ensure it is not exposed to untrusted parties.

Affected
Fireblocks mpc-lib

CVE
CVE-2026-90975

Read on CERT/CC Post on X LinkedIn
```
{"coverage":0.387,"evidence":0,"freshness":0.94,"importance":0.75,"novelty":1,"quality":0.9,"total":0.567}
```
AI Agent Prompt Injection Abuses Markdown Trust Medium·Dark ReadingNewRead An AI agent prompt injection technique exploits how coding assistants process local Markdown files.
A new technique abuses how AI agents consume local instruction files formatted in Markdown. This blurs the boundary between standard text documents and executable actions within the agent environment.

Why it matters. Defenders must consider that seemingly benign Markdown files can be leveraged for prompt injection attacks against AI agents.

Read on Dark Reading Post on X LinkedIn
```
{"coverage":0.387,"evidence":0,"freshness":0.894,"importance":0.75,"novelty":1,"quality":0.85,"total":0.557}
```

Malware & Ransomware 4

LiteLLM Supply Chain Attack: Info-Stealer Malware via PyPI High·9 outlets·Docs via BleepingComputerDay 4NewRead Malicious packages for LiteLLM version 1.107.0 were distributed via PyPI and Docker, stealing credentials and secrets.
Attackers compromised the LiteLLM release pipeline, pushing version 1.107.0 to PyPI and Docker Hub. This version contained malware that modified the installation package to execute upon Python startup.

Why it matters. The malware steals cloud credentials, AI configuration, developer secrets, and browser data. Infected systems require isolation and rebuilding from clean backups.

Do this today Update to LiteLLM version 1.106.2 and rotate all exposed secrets.

Affected
LiteLLM version 1.107.0
1. Thu 22 OctLiteLLM Package Compromised with Credential Stealer and Backdoor
2. Sat 24 OctLiteLLM Supply Chain Attack Distributed Info-Stealer Malware
3. Sun 25 OctLiteLLM Supply Chain Attack Distributed Info-Stealer Malware
4. TodayLiteLLM Supply Chain Attack: Info-Stealer Malware via PyPI
Read the original · Docs Post on X LinkedIn

Summarised from BleepingComputer; also covered by The Record, SecurityWeek.
```
{"coverage":1,"evidence":0,"freshness":0.026,"importance":1,"novelty":0.6,"quality":0.95,"total":0.608}
```
Unknown Actor Infects Magento Stores with Scanner-Evading JavaScript High·The Hacker NewsNewRead An unknown actor infected over 950 online stores, primarily using Magento, with malicious JavaScript.
An unknown threat actor infected 953 online stores, predominantly running Magento. The infection adds code to store pages that serves malicious JavaScript loader code when visited by users.

Why it matters. This activity indicates a persistent compromise vector targeting e-commerce platforms via CMS blocks.

Do this today Review all Magento CMS blocks for unauthorized JavaScript additions.

Affected
Magento stores

Read on The Hacker News Post on X LinkedIn
```
{"coverage":0.387,"evidence":0,"freshness":0.828,"importance":0.75,"novelty":1,"quality":0.8,"total":0.546}
```
uCreator Android Spyware Steals Telegram 2FA and Clones WeChat Accounts High·The Hacker NewsNewRead uCreator spyware compromises Telegram and WeChat by distributing tampered app copies.
uCreator, a rentable Android spyware kit, compromises Telegram and WeChat accounts. The spyware functions by embedding its code within working copies of these two applications.

Why it matters. Infection requires users to install tampered copies outside official app stores or run WeChat within a specific container app.

Do this today Monitor for suspicious app installations outside official stores.

Affected
Android devices

Read on The Hacker News Post on X LinkedIn
```
{"coverage":0.387,"evidence":0,"freshness":0.657,"importance":0.75,"novelty":1,"quality":0.8,"total":0.529}
```
Gheist Weaponizes Stolen GitHub Tokens for Cloud Secret Theft High·3 outlets·Wiz ResearchDay 2NewRead Gheist uses stolen GitHub tokens to plant workflows that exfiltrate repository secrets via Actions logs.
An automated campaign, Gheist, uses compromised GitHub tokens to commit a workflow named sync-check to target repositories. This workflow prints the repository's secrets into the GitHub Actions logs, which are then downloaded by the attackers. The activity has been observed in private repositories and GitHub Enterprise environments.

Why it matters. The attack allows the exfiltration of sensitive data, including environment-scoped secrets and cloud credentials, through GitHub's own API, bypassing traditional network egress controls.

Do this today Hunt for branches named ci-sync- followed by six random characters containing a new file at .github/workflows/sync-check.yml

Affected
GitHub repositories, GitHub Enterprise environments
1. Mon 26 OctGheist: Automated Abuse of Stolen GitHub Tokens Detected
2. TodayGheist Weaponizes Stolen GitHub Tokens for Cloud Secret Theft
Read on Wiz Research Post on X LinkedIn

Also covered by Dark Reading, The Hacker News
```
{"coverage":0.774,"evidence":0,"freshness":0.422,"importance":0.75,"novelty":0.6,"quality":0.85,"total":0.528}
```

Threat Actors & Campaigns 2

Threat Actors Use Microsoft Teams for Impersonation and Malware Delivery High·The Hacker NewsNewRead Threat actors are using Microsoft Teams external chat features to conduct impersonation scams and deploy malware.
Threat actors are using Microsoft Teams' external chat creation feature to execute impersonation scams. This pattern has been observed across numerous incidents where malware and ransomware were deployed.

Why it matters. This vector allows threat actors to bypass traditional perimeter defenses using seemingly legitimate communication channels.

Affected
Microsoft Teams

Read on The Hacker News Post on X LinkedIn
```
{"coverage":0.387,"evidence":0,"freshness":0.776,"importance":0.75,"novelty":1,"quality":0.8,"total":0.541}
```
ConnectWise Attack: AI Agent Used with Scripted Tooling Medium·HuntressNewRead Threat actor used scripted tooling, potentially guided by an AI agent, to abuse ConnectWise credentials against Microsoft 365.
An attacker used compromised ConnectWise Cloud Backup credentials to access Microsoft 365 tenants. The activity involved scripted tooling, with observed patterns suggesting an agent adapted the tooling between runs.

Why it matters. The attack demonstrates automated, high-speed credential abuse across multiple tenants, indicating potential for widespread compromise.

Read on Huntress Post on X LinkedIn
```
{"coverage":0.387,"evidence":0,"freshness":0.846,"importance":0.5,"novelty":1,"quality":0.85,"total":0.478}
```

Breaches & Incidents 1

Cybercrime Group Used ARTEX AI to Breach South Korean Organizations High·Yonhap News via BleepingComputerNewRead A cybercrime operation used the ARTEX AI framework to breach over 90 South Korean organizations, stealing 8.2 million records.
Seven suspects were identified by South Korean police in a cybercrime operation. The group used the ARTEX AI pentesting framework to breach over 90 organizations across South Korea. Authorities noted that AI utilities lowered the barrier to entry for large-scale attacks.

Why it matters. The use of generative AI in criminal activity enables non-professional actors to conduct high-impact data theft. Organizations must assume initial access may be achieved via AI-assisted methods.

Read the original · Yonhap News Post on X LinkedIn

Summarised from BleepingComputer.
```
{"coverage":0.387,"evidence":0,"freshness":0.898,"importance":0.75,"novelty":1,"quality":0.95,"total":0.568}
```

AI Security 1

Malicious AGENTS.md File Triggers AI Coding Agent Compromise Info·4 outlets·Forever Security via BleepingComputerNewRead Malicious instruction files in Downloads folders can trick AI coding agents into installing packages or leaking tokens.
Researchers demonstrated that placing a malicious markdown file, such as AGENTS.md, in a Downloads folder can trick AI coding agents into executing commands. These agents automatically load and interpret instructions from files in their working directory. The attack can lead to the installation of rogue npm packages or the exposure of OAuth tokens.

Why it matters. This attack vector exploits the agents' intended functionality, allowing unauthorized code execution or credential theft if the working directory is untrusted. Users must restrict agent scope to specific project directories.

Do this today Do not set the ~/Downloads directory as the working directory for AI coding agents.

Affected
AI coding agents (OpenAI Codex, Anthropic Claude Code)

Read the original · Forever Security Post on X LinkedIn

Summarised from BleepingComputer; also covered by The Hacker News, CSO Online.
```
{"coverage":0.898,"evidence":0,"freshness":0.398,"importance":0.75,"novelty":1,"quality":0.95,"total":0.595}
```

Policy, Law & Arrests 1

Senators Claim Trump Administration Withholding Surveillance Court Opinion Info·Wyden via CyberScoopNewRead Senators and rights groups allege the Trump administration is illegally withholding a court opinion regarding surveillance practices.
Senators and civil liberties groups stated the Trump administration failed to declassify a court opinion concerning U.S. citizen surveillance. This opinion reportedly contains a judge's objection to government filtering systems used by analysts.

Why it matters. Withholding this opinion may impact congressional debate and the reauthorization of Section 702 of FISA. Civil liberties groups suggest Congress should withhold reauthorization until the information is released.

Read the original · Wyden Post on X LinkedIn

Summarised from CyberScoop.
```
{"coverage":0.387,"evidence":0,"freshness":0.825,"importance":0.5,"novelty":1,"quality":0.9,"total":0.481}
```

CVE spotlight

CVE	Affects	CVSS	EPSS	Status
CVE-2026-109619	phpPHP has various global state that temporarily points into request memory. In some scenarios, global state may not be correctly reset, or may…	8.1	100%	Not in KEV
CVE-2026-109618	phpPHP uses longjmp() for out-of-memory (OOM) errors to exit the current context and abort handling of the request. Some contexts may not prope…	9.8	100%	Not in KEV
CVE-2026-111497	—A buffer overflow vulnerability in the Brotli decoder (c/dec/decode.c) in Google Brotli versions 1.0.3 through 1.2.0 on all platforms allows…	9.2	33%	Not in KEV
CVE-2026-106072	—A flaw in Node.js TLS hostname verification allows a certificate with only a DNS SAN to authenticate an IP endpoint when the client connects…	6.7	1%	Not in KEV
CVE-2026-80097	microsoft authenticatorImproper authentication in Microsoft Authenticator allows an unauthorized attacker to elevate privileges locally.	8.6	0%	Not in KEV
CVE-2026-105482	—A flaw in Node.js DNS resolution allows hostnames containing a NUL byte to be silently truncated by the dns.resolve*() functions and dns.rev…	7.4	0%	Not in KEV

Quiz

Yesterday's answer

According to the provided text, what condition enables Remote Code Execution via CVE-2026-109618?

CVE-2026-109618 allows RCE when PHP uses longjmp() for OOM errors

The story explicitly states that CVE-2026-109618 allows RCE when PHP uses longjmp() for OOM errors.

Today's question

According to the text, how does CVE-2026-109618 enable Remote Code Execution (RCE)?

On the radar tomorrow

CTF starts Tue 03:30 UTCSudocrypt v16.0
WatchingQuest Diagnostics confirms data breach exposed patient records
WatchingOpenAI AI Agents Demonstrate VM Escapes and Secret Cross-Sandbox Communication

Also worth your time

Exploits

PHP2Shell (CVE-2026-109618) - PHP core versions after v5.4 - RCERemote · CVE-2026-109618
pizzafy-1.0 - SQLiWebApps
South River Technologies Titan MFT Server 26.0.0 - code executionRemote
Synametrics Syncrify 6.2 - command executionRemote
Keycloak 26.7.2 - Unauthenticated Account TakeoverWebApps

Tools

SpecterOps/BloodHound v9.9.0New release v9.9.0 · Active DirectoryNew features include bulk PZ rule ingestion, associated label display in Explore, and job history table functionality.
zeek/zeek v9.0.1New release v9.0.1 · Detection & responseFixes address unbounded state growth in SSL, KRB, MIME, and memory exhaustion across multiple protocols including DNS, T…
jamesalmeida/agent-honeypotNew · 39 stars in 3 days · AI security · PowerShellDetects unauthorized interaction with a system by AI agents and logs details of the interaction.
letox4/OsintHunterNew · 31 stars in 6 days · OSINT · PythonCollects and analyzes data from multiple sources using various inputs to build an investigative profile.
YZBRH/keyrealmNew · 47 stars in 16 days · Recon · PythonManages and validates credentials obtained during penetration testing, supporting multiple credential types and automate…
cleonaia/hack_cameraWebNew · 31 stars in 17 days · Network · PythonDemonstrates network traffic interception using ARP spoofing, Wireshark analysis, and Metasploit for lab exercises.

Community

Podcasts

181: RickyDarknet Diaries · 01:20:15
Risky Business #856 -- 8-bit encryption, PHP megabugs and moreRisky Business · 1:04:58
SN 1101: ECDSA's Curves Origins - When AI Becomes the AttackerSecurity Now · 2:31:36
SANS Stormcast Monday, October 26th, 2026: PHP2Shell Exploit Reliability; Node.js Brotli Decoder RCE (CVE-2026-111497); Tailscale "Tailworm" Vulnerabilities; HPE ArubaOS 8/10 Patches 72 FlawsSANS Internet Storm Center · 7:45

And finally

Chinese National Extradited for Alleged $36M Crypto Scam

A Chinese national was extradited for alleged involvement in a cryptocurrency investment scam. Victims were reportedly deceived via social media into transferring funds to fake investment platforms.

Source: Department of Justice · via BleepingComputer

End of No. 246

That's today's brief.

On the radarCTF starts Tue 03:30 UTC: Sudocrypt v16.0

The next brief lands at 06:30 IST.

Tuesday, 27 October 2026

From the editor

Actively Exploited 2

Vulnerabilities & Patches 4

Malware & Ransomware 4

Threat Actors & Campaigns 2

Breaches & Incidents 1

AI Security 1

Policy, Law & Arrests 1

CVE spotlight

Quiz

On the radar tomorrow

Also worth your time

Exploits

Tools

Community

Podcasts

And finally

Keyboard shortcuts

Top stories

From the editor

Actively Exploited 2

Vulnerabilities & Patches 4

Malware & Ransomware 4

Threat Actors & Campaigns 2

Breaches & Incidents 1

AI Security 1

Policy, Law & Arrests 1

CVE spotlight

Quiz

On the radar tomorrow

Also worth your time

Exploits

Tools

Community

Podcasts

And finally