Daily cybersecurity briefing

Top 10 Cybersecurity Stories for October 26, 2026

Researchers confirmed in-the-wild PHP2Shell remote code execution against an Apache mod_php honeypot running PHP 8.4.26; prioritize the released patches for CVE-2026-109618 and CVE-2026-109619, and apply PHP’s recommended session.upload_progress.enabled=0 mitigation for the separate, unpatched CVE-2026-115560. Wiz also reports successful Gheist secret theft through malicious GitHub Actions workflows in private repositories and Enterprise environments, warranting affected-secret rotation and pushing-token revocation, while Brotli decoder updates require product-by-product deployment and service restarts despite no reported real-world exploitation.

Compiled by the Slugnet Editorial System. Published Oct 26, 2026, 7:41 AM EDT

Audio briefing

Listen to this edition

A spoken version of today’s prioritized cybersecurity briefing.

  1. 01
    BleepingComputer

    PHP2Shell security flaws now sucessfully exploited in the wild

    What happened

    Researchers confirmed in-the-wild exploitation of PHP2Shell, the remote code execution flaws CVE-2026-109618 and CVE-2026-109619, against an Apache mod_php honeypot running PHP 8.4.26. Captured traffic showed an 800 KiB memory disclosure followed by an execution payload using a function address derived from that data, with a shell command running immediately afterward. The PHP security team released patches on October 21; earlier observed attempts had been reported as unsuccessful.

    Why it ranks #1

    Researchers confirmed PHP2Shell exploitation against an Apache mod_php honeypot, capturing memory disclosure followed by command execution; the PHP security team released patches on October 21.

    Who should care

    CISOs and security leaders, SOC and incident response teams

    What to do

    Set session.upload_progress.enabled=0 to block potential attacks, as PHP developers advised before patches for the PHP2Shell security updates were released.

    Impact
    high
    Urgency
    immediate
    Confidence
    high
    Scope
    enterprise
    Status
    confirmed incident
    Read the original source Link to this ranking Share on Bluesky Share by email
  2. 02
    The Hacker News

    Gheist Uses Stolen GitHub Tokens to Steal Private Repository Secrets via Actions Logs

    What happened

    Wiz identified an automated campaign, dubbed Gheist, that uses stolen GitHub tokens to plant secret-stealing Actions workflows, recovering more than 13,500 planted branches across over 850 compromised accounts. The workflows encode accessible secrets in base32 to evade log masking; attackers then download the logs through GitHub’s API and delete the runs. Wiz reported successful attacks in private repositories and GitHub Enterprise environments, while GitHub’s suspicious-workflow approval safeguard stopped every public-repository run it examined; that safeguard does not cover private repositories or GitHub Enterprise Server.

    Why it ranks #2

    Gheist used stolen GitHub tokens to steal secrets through Actions logs in private repositories and GitHub Enterprise environments, outside the approval safeguard that blocked examined public-repository runs.

    Who should care

    IT and platform operations, SOC and incident response teams

    Impact
    critical
    Urgency
    near-term
    Confidence
    high
    Scope
    enterprise
    Status
    confirmed incident
    Read the original source Link to this ranking Share on Bluesky Share by email
  3. 03
    BleepingComputer

    AI coding agents hijacked using AGENTS.md file in Downloads folder

    What happened

    Forever Security traced an AI coding agent’s attempt to install the malicious npm package function-color on a customer’s computer to an AGENTS.md file in ~/Downloads, which was also the agent’s working directory; how the file arrived remains unknown. The agent loaded the file as project instructions, allowing attacker-controlled text to direct its tools. Forever Security researchers reported that a reconstructed attack against Codex opened the calculator app in 90% of tests, while a separate test using a malicious CLAUDE.md file redirected Claude Code’s API traffic to a researcher-controlled server, exposing its OAuth token and allowing the server to return tool calls that executed code in roughly 50% of tests.

    Why it ranks #3

    An AI coding agent attempted to install a malicious npm package after treating an AGENTS.md file in its Downloads working directory as project instructions.

    Who should care

    CISOs and security leaders, SOC and incident response teams

    What to do

    Avoid setting ~/Downloads as an AI coding agent’s working directory because it may contain untrusted files that the agent could interpret as instructions.

    Impact
    high
    Urgency
    near-term
    Confidence
    high
    Scope
    enterprise
    Status
    confirmed incident
    Read the original source Link to this ranking Share on Bluesky Share by email
  4. 04
    The Hacker News

    uCreator Android Spyware Can Steal Telegram 2FA Passwords and Clone WeChat Accounts

    What happened

    Researcher Ovi analyzed uCreator, a rentable Android spyware kit that embeds malicious code in tampered Telegram and WeChat apps distributed outside official stores, or runs WeChat inside an attacker-controlled container. The code can capture Telegram’s two-step verification password before hashing, query chat databases and hide attacker sessions; its WeChat component can export account data for cloning. These findings establish technical capabilities, not confirmed end-to-end account takeovers or a reliable current victim count.

    Why it ranks #4

    Researcher Ovi found uCreator spyware can capture Telegram verification passwords and export WeChat account data for cloning through tampered Android apps, though completed takeovers remain unconfirmed.

    Who should care

    Individual users, Identity and access teams, IT and platform operations, SOC and incident response teams

    Impact
    high
    Urgency
    monitor
    Confidence
    high
    Scope
    consumer
    Status
    emerging
    Read the original source Link to this ranking Share on Bluesky Share by email
  5. 05
    The Hacker News

    Brotli Decoder Flaw Behind Node.js RCE Also Affects Bun, Python and Other Widely Used Software

    What happened

    CVE-2026-111497, a memory-corruption flaw in Google’s Brotli decoder, affects software including Node.js, Bun, .NET, Python packages, Envoy, FFmpeg, ClickHouse and Apache Arrow when they decompress attacker-controlled data. Imperva demonstrated code execution on Node.js 24.22.0 for Linux ARM64, while JFrog demonstrated denial of service in ClickHouse and pyarrow; no in-the-wild exploitation has been reported. Fixes are arriving product by product, including Microsoft’s October 25 .NET updates, because affected applications may bundle separate copies of the decoder.

    Why it ranks #5

    A Brotli decoder flaw affects multiple products processing attacker-controlled data, with demonstrated Node.js code execution and product-specific fixes required because applications may bundle separate decoder copies.

    Who should care

    Application security teams, IT and platform operations, SOC and incident response teams

    Impact
    high
    Urgency
    monitor
    Confidence
    high
    Scope
    product
    Status
    disclosed
    Read the original source Link to this ranking Share on Bluesky Share by email
  6. 06
    The Hacker News

    Unpatched Tailscale SSH Flaws Could Let a Hacked Device Reach Root on Other Servers

    What happened

    Erica Windisch of Cyberdione Labs reported lab tests showing that an attacker controlling an authorized device could gain root on reachable servers through Tailscale SSH port forwarding, testing Tailscale 1.104.2 and headscale 0.29.5. The reported mechanism causes the root-level Tailscale process to access privileged system sockets; it requires an existing foothold, an SSH access grant and forwarding enabled. No real-world attacks or independent reproductions were reported, and Tailscale acknowledged issues under investigation without confirming the specific claims; no fixed version was available at publication.

    Why it ranks #6

    Reported lab tests showed Tailscale SSH forwarding could enable root access from an authorized compromised device; exploitation requires SSH access and forwarding, with no fix available at publication.

    Who should care

    CISOs and security leaders, SOC and incident response teams

    Impact
    moderate
    Urgency
    near-term
    Confidence
    high
    Scope
    enterprise
    Status
    confirmed incident
    Read the original source Link to this ranking Share on Bluesky Share by email
  7. 07
    Help Net Security

    Forgotten AI agents may still access sensitive company data

    What happened

    Arrakis Security’s 2026 Autonomous Agents Threat Report found that 76% of enabled agents examined had not run in 60 days but retained credentials and connections to company systems. About a quarter of dormant agents could still be activated externally through unsigned webhooks, allowing them to run with their stored access. Arrakis Security’s analysis of thousands of agents in customer production environments over 12 weeks found that about 76% of enabled agents had not run in 60 days but retained credentials and connections to company systems.

    Why it ranks #7

    Arrakis Security found that 76% of examined enabled agents were dormant but retained company-system access, with about a quarter externally activatable through unsigned webhooks.

    Who should care

    CISOs and security leaders

    Impact
    moderate
    Urgency
    monitor
    Confidence
    high
    Scope
    enterprise
    Status
    emerging
    Read the original source Link to this ranking Share on Bluesky Share by email
  8. 08
    Help Net Security

    AI slop is piling up highest on the biggest open-source projects

    What happened

    In a Linux Foundation survey of 234 open-source contributors and maintainers, respondents from projects with more than 50 contributors estimated that nearly a third of incoming code changes were low quality or AI-generated. Large projects reported the fastest growth in submissions and the most complaints about review delays, with maintainer burnout their leading concern about AI-assisted contributions. About a third of projects formally allow AI-assisted contributions, generally requiring disclosure and contributor review, while three in four maintainers use AI tools to help assess submissions.

    Why it ranks #8

    A Linux Foundation survey found large open-source projects reporting growing submission volumes and review delays, with maintainer burnout their leading concern about AI-assisted contributions.

    Who should care

    Application security teams, IT and platform operations, SOC and incident response teams

    Impact
    moderate
    Urgency
    monitor
    Confidence
    high
    Scope
    product
    Status
    disclosed
    Read the original source Link to this ranking Share on Bluesky Share by email
  9. 09
    Help Net Security

    Synthetic identities leave US lenders with $2.4 billion in fraud exposure

    What happened

    TransUnion estimates that US lenders faced $2.4 billion in potential losses from synthetic identity fraud at the end of June 2026, including $1.5 billion in auto lending; these figures represent exposure, not losses already incurred. Its report describes criminals combining breached personal information with AI-generated documents to create identities that pass onboarding checks and obtain credit across multiple services. Separately, 81% of suspected digital fraud recorded in TransUnion’s network during the first half of 2026 occurred at account login or during financial transactions, rather than account creation.

    Why it ranks #9

    TransUnion estimates $2.4 billion in synthetic identity fraud exposure for US lenders, with criminals using breached personal information and AI-generated documents to pass onboarding checks.

    Who should care

    CISOs and security leaders, Identity and access teams, IT and platform operations, SOC and incident response teams

    Impact
    moderate
    Urgency
    monitor
    Confidence
    high
    Scope
    enterprise
    Status
    confirmed incident
    Read the original source Link to this ranking Share on Bluesky Share by email
  10. 10
    Help Net Security

    Anthropic details what stops Claude Duet from going rogue

    What happened

    Anthropic outlined security controls for Claude Duet, its research-preview agent that monitors connected accounts and performs tasks while users are away. The company says approval gates and a separate action check operate outside the task-performing model, while credential-handling infrastructure keeps passwords, cookies and sign-in tokens outside the agent’s reach. A separate classifier checks outgoing messages and uploads for sensitive information, but Anthropic acknowledges that these checks may miss sensitive content or flag harmless messages.

    Why it ranks #10

    Anthropic outlined external approval gates, credential isolation, and outbound checks for Claude Duet, providing control details while acknowledging that sensitive-information checks can miss content or flag harmless messages.

    Who should care

    CISOs and security leaders

    Impact
    moderate
    Urgency
    monitor
    Confidence
    high
    Scope
    enterprise
    Status
    emerging
    Read the original source Link to this ranking Share on Bluesky Share by email