Security news, decoded.

Microsoft, working with Resecurity, has exposed and dismantled an underground marketplace called YAM Store (tracked as Storm-3119) that sold phishing kits and stolen cloud accounts. Three kits — CoGUI, Haozi and Iron Man System — built by a group Microsoft labels Storm-2990, were used to send over a billion phishing emails and SMS messages since January 2026, hitting more than 75,000 organizations in healthcare, finance, wholesale distribution and IT across Japan, the US, Australia and India. Beyond phishing templates impersonating banks, retailers and government agencies in 139+ countries, the store also sold fraudulently obtained Microsoft and Google cloud accounts, repurposing compromised SharePoint and OneDrive storage as hosting infrastructure for buyers. Following a tip from Japan's police, Microsoft and Health-ISAC obtained a court order from the Southern District of Florida to seize domains, disabling over 20,000 linked accounts and blocking more than 100,000 fraud-related domains.

Security firm Previdian reports active exploitation of a critical Langflow flaw tracked as CVE-2026-105740 (CVSS 9.9), which lets any logged-in user, even a non-admin account, run arbitrary system commands on the server. The bug stems from Langflow passing an attacker-controlled command field straight into a shell call when a malicious MCP server with 'stdio' transport is registered, with no validation, allowlisting or sandboxing, and it also permits injecting arbitrary environment variables. Previdian's honeypots logged 1,423 exploitation attempts from 27 distinct IP addresses across more than a dozen countries since September 19, 2026, with activity spiking in mid-October, and observed attackers harvesting secrets like LANGFLOWSECRETKEY, probing Docker and host files, and reusing one IP already tied to attacks on Flowise and other Langflow CVEs. Langflow, now owned by IBM via DataStax, urges users on versions before 1.9.0 to update immediately, disable auto-login, enforce strong unique passwords, and monitor MCP server endpoint logs for suspicious requests.

Palo Alto Networks has shipped fixes for a zero-day vulnerability in PAN-OS that attackers were already abusing in limited real-world attacks. The high-severity bug, CVE-2026-0322, resides in the management interface and lets an authenticated attacker with network access run arbitrary code with root privileges on affected firewalls, with no special configuration required to trigger it. It impacts PAN-OS running on PA-Series and VM-Series firewalls as well as on Panorama, and was found by the vendor's own internal research rather than an outside reporter. Palo Alto has published a handful of related IP indicators of compromise, though details on the attackers or their methods remain undisclosed, and CISA has already added the flaw to its Known Exploited Vulnerabilities catalog, giving federal agencies until October 17 to remediate. The same day, Palo Alto also released about a dozen advisories covering lower-severity issues across PAN-OS, GlobalProtect and Prisma Cloud, and refreshed Prisma Browser with current Chromium security updates.

Ahsay Backup has issued fresh hotfixes for its AhsayCBS product after discovering that August 2026 patches for two vulnerabilities, CVE-2026-105133 and CVE-2026-105134, were incomplete. Attackers had been chaining these two bugs to bypass authentication and gain unauthenticated remote code execution, then planting web shells and XMRig cryptocurrency miners on exposed servers, as detailed in earlier reporting from Huntress. The vendor now says the real fix is delivered in AhsayCBS v10.3.4.45 and v9.15.2.70, and is urging customers to apply the hotfix and reboot their servers immediately; Huntress has confirmed the v10.3.4.45 build closes the initial entry point that remained open in prior v10 releases. Because the flaws are being actively exploited, CISA has added them to its Known Exploited Vulnerabilities catalog, giving federal agencies until October 17, 2026 to remediate.

Cisco has disclosed that attackers are actively exploiting a critical flaw, CVE-2026-76519 (CVSS 9.8), affecting the Advanced Malware Protection component of its Secure Email Gateway appliance. The bug lets unauthenticated remote attackers send malicious email attachments that write arbitrary files to the underlying operating system, ultimately enabling root-level command execution, when file reputation scanning is enabled with Content Scanner Tools versions older than 25.3.0.5525. Cisco says its PSIRT first observed exploitation in October 2026 and has already contacted customers whose Cisco Secure Email Cloud devices showed signs of compromise, while rolling out mitigations it can control directly. The company had actually shipped fixes back on September 14 in AsyncOS versions 15.5.5-014 and 16.5.0-780, originally bundled under the broader advisory for CVE-2026-76440, before updating that advisory this week to flag in-the-wild attacks. Administrators are urged to inspect external logs for unusual uploads or downloads, contact Cisco TAC if compromise is suspected on physical hardware, and for virtual appliances to rebuild from scratch on patched software while rotating credentials. This marks the second Secure Email Gateway zero-day exploited since mid-September, following the similarly root-granting CVE-2026-76461, and CISA has added CVE-2026-76519 to its Known Exploited Vulnerabilities catalog with a federal remediation deadline of October 17.

Palo Alto Networks has released fixes for a high-severity authenticated remote code execution bug in PAN-OS, identified as CVE-2026-0322 (CVSS 8.6), after confirming it is being actively abused in the wild. The flaw sits in the management interface and lets an attacker who already has network access and valid credentials run arbitrary code with root privileges on the firewall, with no special setup needed to trigger it. It hits PAN-OS running on PA-Series and VM-Series firewalls plus Panorama appliances (virtual and M-Series), while Cloud NGFW and Prisma Access remain unaffected. The vendor says it has only observed very limited attack activity so far, tied to a handful of US-based IP addresses, and recommends locking down management-interface access to trusted hosts or a jump box while applying the patched PAN-OS builds. CISA has added this flaw, alongside a separate critical Cisco Secure Email Gateway bug (CVE-2026-76519), to its Known Exploited Vulnerabilities catalog, giving federal agencies until October 17, 2026 to remediate.

Palo Alto Networks' Unit 42 has uncovered a cyber-espionage operation, tracked as CL-UNK-1211, that struck government bodies and state-linked Linux systems between October 2025 and April 2026. Researchers found that many of the intrusion tools carried conversational comments and reasoning notes typical of large-language-model output, indicating the operators used AI to generate or even run parts of their malware automatically. The attackers combined open-source utilities, a custom cross-platform backdoor called SysGoRAT, and numerous web shells to gain persistent footholds, dump databases, and raid government mailboxes in search of linked social-media credentials. To cover their tracks they tried to disable security agents, wipe logs, falsify file timestamps, and plant fake red-team artifacts, though Palo Alto Networks says its Cortex XDR platform detected and blocked the key evasion attempts.

Palo Alto Networks has patched a high-severity zero-day in PAN-OS, identified as CVE-2026-0322, after confirming it was already being abused against firewalls in the field. The flaw sits in the PAN-OS management interface and allows an authenticated attacker with network access to run commands with root privileges, carrying a severity rating of 8.6. Affected gear includes PA-Series and VM-Series firewalls plus Panorama virtual and M-Series appliances, while Cloud NGFW and Prisma Access remain unaffected. The vendor says exploitation so far has been limited, has released fixes for PAN-OS 10.2, 11.1, 11.2, 12.1, and 12.2 (with the 11.2.16 build due October 15), and urges customers to restrict management-interface access to trusted internal IPs since no workaround exists. CISA has added the bug to its Known Exploited Vulnerabilities list, giving federal agencies until October 17 to apply mitigations, and Palo Alto has published three attacker IP addresses tied to the activity.

Ivanti disclosed CVE-2026-104935, a combined authentication-bypass and unauthenticated remote-code-execution flaw in Ivanti Connect Secure SSL-VPN gateways, after confirming active exploitation against a limited set of customers. Notably, Ivanti had already found part of this bug back in July through its AI-assisted vulnerability-hunting program and fixed it in ICS 25.1.3.0, but attackers struck while the company was still backporting the patch to older supported versions. Researchers at Volexity attribute the attacks to a group tracked as UTA0576, which chained the bug with an as-yet-unpatched privilege-escalation flaw to gain root-level code execution on fully updated appliances, harvest credentials, and move into Active Directory before being detected. CISA has added CVE-2026-104935 to its Known Exploited Vulnerabilities catalog, giving federal agencies until October 16 to patch or disable affected devices, while experts warn the 8.6 CVSS score understates the real-world risk given the internet-facing exposure and active in-the-wild exploitation.

Cisco has confirmed that attackers are actively exploiting a critical, unauthenticated vulnerability in its Secure Email Gateway, tracked as CVE-2026-76519 with a CVSS score of 9.8. The bug lets a remote sender, without needing any credentials, deliver a specially crafted attachment that writes files to the device's operating system, potentially enabling root-level command execution. It only affects gateways that have the File Reputation (AMP) service turned on and are running a Content Scanner Tools package older than version 25.3.0.5525; Cisco discovered the issue through its own internal scanning and learned of real-world attacks in October. Fixes were actually already included in the September 14 hardening builds AsyncOS 15.5.5-014 and 16.5.0-780, plus the standalone scanner update, and this marks the second Secure Email Gateway bug — after the SQL-injection flaw CVE-2026-76461 disclosed the same month — that Cisco says is being actively exploited.

Security researchers report active exploitation of a critical flaw in Collabora's optional "Built-in CODE Server" app (richdocumentscode) used by Nextcloud Office, tracked as CVE-2026-110447. The bug lets an unauthenticated attacker inject shell commands through the app's proxy.php script whenever its local document server isn't running—such as during a restart, upgrade, or crash—allowing a web shell to be dropped on the Nextcloud host. At least three administrators confirmed compromises starting September 10, with attackers in one case searching for cryptocurrency wallets and credential files like kdbx and wallet.dat. Collabora patched the issue in version 26.4.303 on September 11 but only published the advisory on October 14; simply disabling the app does not mitigate the flaw, so affected installations must upgrade or remove richdocumentscode entirely, with older Nextcloud releases (29 and earlier) requiring a full platform upgrade.

Cisco has released a fix for another zero-day vulnerability in its Secure Email Gateway (SEG) that was already being exploited by attackers before the patch was available. Tracked as CVE-2026-76519, the bug is a relative path traversal flaw in the Advanced Malware Protection (AMP) attachment-handling feature, which lets an unauthenticated remote attacker send a crafted email attachment to write arbitrary files and run commands with root privileges on the appliance's underlying OS. The flaw only affects devices with File Reputation enabled on a mail policy and running Content Scanner Tools versions older than 25.3.0.5525; Cisco published log-based indicators of compromise and urged admins to check firewall logs for signs of attacker cleanup activity. Cisco says it has already mitigated its own cloud instances and reached out to customers with suspected compromises, recommending credential and certificate renewal, and this marks the latest in a string of actively exploited SEG/AsyncOS zero-days disclosed by Cisco this year, following CVE-2026-76461 and CVE-2025-20393.

This month's ICS Patch Tuesday brought a heavy round of disclosures from Siemens, Schneider Electric, and Rockwell Automation. Siemens released 17 advisories, fixing critical bugs in certain Simatic S7-1500 CPUs (over 400 flaws tied to the GNU/Linux subsystem), Industrial Edge devices, and the Ruggedcom Crossbow Station Access Controller, alongside high- and medium-severity issues across products like Simatic IPC, WinCC, Desigo, Talon, and Capital Embedded AR Classic. Schneider Electric disclosed two high-severity bugs affecting Pro-face STW6000/Harmony STW6 (device takeover via a malicious webpage) and Sage RTUs (malicious firmware installation by a privileged user). Rockwell Automation's advisory covers two 1732E ArmorBlock vulnerabilities, including a critical flaw letting an unauthenticated attacker reset the web server password and seize admin control, plus a denial-of-service issue. CISA also published advisories for Swann SWIFI-SE2KIC cameras and an Ignition privilege-escalation bug, while Germany's CERT@VDE flagged an information-disclosure issue in Bender's Customer&Partner platform; ABB separately noted anomalous password-reset activity tied to a dynovaPRO credential-reset bypass (CVE-2026-18963) affecting nine customers.

Cisco Talos, working with Recorded Future's Insikt Group, has exposed a Kazakhstan-linked threat cluster dubbed UAT-6275 (also tracked as TAG-193 and overlapping with Silent Lynx) that has been spying on government bodies across Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Uzbekistan, Armenia, Azerbaijan, Russia and Belarus since June 2025, with particular interest in diplomacy, water resources and energy. Between June 2025 and August 2026 the actor used four separate delivery chains — malicious macro documents, PS2EXE-built downloaders, GitHub-hosted PowerShell scripts, and a weaponized WinRAR archive abusing CVE-2025-8088 — to drop custom reverse shells such as PickleShell, Chvost and Laplas for interactive remote access. Researchers also found a public proof-of-concept for the FortiWeb flaw CVE-2025-25257 and Linux-capable tools staged on exposed infrastructure, hinting at plans to broaden targeting beyond Windows. Talos recommends monitoring suspicious PowerShell and startup-folder activity and promptly patching archive software to blunt this exploit-driven delivery method.

CISA has flagged an Ivanti Connect Secure (ICS) authentication bypass, CVE-2026-104935, as actively exploited since at least October 1, giving federal agencies until Friday to secure their systems under Binding Operational Directive 26-04. The high-severity flaw allows an unauthenticated attacker to run arbitrary commands remotely without any user interaction, and Ivanti shipped fixes in versions 22.8R2.5, 25.1.3.0, 25.1.3.1, and 25.1.1.2 while urging customers to patch immediately. Security firm Volexity says a threat group it calls UTA0576 has been chaining this bug with two additional vulnerabilities since October 3 to gain unauthenticated root access, then stealing Active Directory credentials, LAPS passwords, and BitLocker recovery keys. Shadowserver counts more than 5,600 internet-facing ICS devices, and CISA warns that Ivanti's Integrity Checker Tool may miss signs of compromise since attacker commands run only in memory, recommending a factory reset, credential rotation, and certificate reissuance for affected appliances.

CISA has issued an emergency directive telling organizations to immediately patch Ivanti Connect Secure (ICS) appliances after confirming active exploitation of a newly disclosed zero-day. The flaw, CVE-2026-104935 (CVSS 8.6), is an authentication bypass affecting ICS versions 25.1.1.1 and earlier, 25.1.2.2 and earlier, and 22.8R2.4 and earlier, letting unauthenticated remote attackers run arbitrary commands, steal credentials, and exfiltrate data. The bug has been added to CISA's Known Exploited Vulnerabilities catalog, with federal agencies required to remediate by October 16, 2026, and the agency warns version 22.8 reaches end-of-life on January 31, 2027. Security firm Previdian reported 105 exploitation attempts against its honeypots as of October 14, 2026, originating from 50 distinct IPs tied to Singapore and the U.S., with attackers abusing an X-DSCGI-PATH header to upload disguised Perl scripts and probe for code execution. Separately, Volexity says it detected exploitation as early as October 3, 2026, by a financially motivated actor named UTA0576, who chained the bug with a privilege-escalation flaw to gain root access, harvest credentials, run Kerberoasting and LDAP reconnaissance tools, and erase forensic traces of the intrusion.

A use-after-free bug in curl's HTTP/2 upload handling, tracked as CVE-2026-92392, has been fixed in curl 8.23.0 after a researcher built a proof of concept showing it can lead to code execution on the uploading client. The flaw, present since version 8.12.0 (February 2025) through 8.22.0, is triggered when a malicious server stalls an HTTP/2 upload, resets the stream, and reuses stale memory tied to a freed transfer handle, allowing an attacker who knows the client's exact build and memory layout to hijack a function pointer. curl rates the issue High—its first High-severity flaw since 2023's SOCKS5 heap overflow—though it notes a real-world attempt would most likely just crash the client rather than achieve execution. Users should upgrade to curl/libcurl 8.23.0 (or the matching 8.14.3, 8.16.2, 8.20.2 branch releases), or disable HTTP/2 via --http1.1 or CURLOPTHTTPVERSION as a stopgap; the release also resolves 23 additional Low-severity CVEs, and curl credits the discovery to a researcher using Anthropic's new opt-in OSS Scanner service.

Security firm XBOW used OpenAI's publicly available GPT-5.6 Sol model to drive AI agents that assembled a complete Chrome exploit chain in only six days, earning Google's maximum $250,000 bug bounty for a full-chain report. The attack combined three previously unknown flaws — a V8 engine issue tracked as CVE-2026-85045, a bug in Chrome's animated-WebP image decoder, and a Crashpad crash-handler issue tracked as CVE-2026-78952 — to escape both of Chrome's sandbox layers and run arbitrary code on a Windows machine simply by visiting a malicious page, no clicks or downloads needed. Google has already shipped fixes in Chrome 152.0.7977.65 and 152.0.7977.82, and XBOW states the bugs were never exploited against real users, though an estimated billion Windows Chrome users were exposed before the patch. The case highlights how AI-assisted research can compress months of exploit-chain development into days, underscoring the need for users to restart their browsers to apply the latest security update.

Security researchers at Volexity have uncovered a threat actor tracked as UTA0576 abusing three previously unknown flaws in Ivanti Connect Secure to breach a fully patched VPN appliance (version 22.8R2.4). By chaining an authentication bypass in welcome.cgi, a file-write bug in wu.cgi, and a privilege-escalation issue in how the dsinvoked broker trusts DSCache data, the attacker obtained root-level command execution. This allowed theft of stored credentials, Active Directory metadata, local admin passwords, BitLocker keys, LDAP traffic, and configuration files, along with log tampering to erase traces of the intrusion. Ivanti has shipped a fix for the authentication-bypass/RCE issue under CVE-2026-104935 and is still developing a patch for the privilege-escalation component; CISA has added the flaw to its Known Exploited Vulnerabilities list and given federal agencies until October 16, 2026 to remediate and investigate, while Volexity warns that the standard Integrity Checker Tool may fail to detect this particular compromise.

Microsoft's latest Patch Tuesday round closed more than 1,400 vulnerabilities, setting a new record, and among them is CVE-2026-96951, a Windows Shell command injection flaw already being abused by attackers. Exploiting it requires getting a victim to open a specially crafted file—delivered through email, chat apps, or a malicious download—after which the file bypasses the usual Mark of the Web safeguards like security warnings, Protected View, or outright blocking. Tenable researcher Satnam Narang pointed out this marks the third Windows Shell bug exploited in the wild this year, following CVE-2026-21510 (tied to APT28) and CVE-2026-32202, with it still unclear whether CVE-2026-96951 stems from an incomplete fix for those earlier issues or represents fresh APT28 activity. Separately, ZDI's Dustin Childs highlighted CVE-2026-102866, a critical signature-verification weakness in Microsoft Office SharePoint that lets unauthenticated attackers hijack user accounts by tampering with a stolen proof token, alongside 55 wormable, zero-click remote-code-execution bugs across components such as DHCP Server, Remote Desktop Gateway, and Network Policy Server that administrators are urged to patch promptly.

Security researchers at Tencent's Zhuque Lab, using their Corvus AI-assisted analysis pipeline, uncovered a nearly decade-old Linux kernel bug in the vhost subsystem, tracked as CVE-2026-74580 and dubbed Vghost. The flaw lets a local user who can open /dev/vhost-net or /dev/vhost-vsock read and write arbitrary kernel memory, bypass KASLR, and escalate to root, and in a second scenario involving virtio behind a virtual IOMMU with device-IOTLB mode, lets a malicious guest VM corrupt the host's QEMU process memory and potentially hijack its execution. The root cause is a stale address-translation cache that isn't invalidated when ring addresses or size are changed via VHOSTSETVRINGADDR or VHOSTSETVRINGNUM while a queue is active, a gap introduced by kernel changes from 2017 and 2020. Affected kernels range from 4.11 through 7.2-rc4, with confirmed exploitation on Ubuntu, RHEL, Amazon Linux, Microsoft Azure Linux, and Oracle Linux; a fix (commit de845981da67) has landed upstream and in several vendor kernels, and administrators should update immediately or restrict access to the vhost device nodes in the meantime.

Citrix, Fortinet, and Ivanti issued coordinated security updates this week addressing a combined total of 29 vulnerabilities, several rated critical. Citrix's three advisories cover nine flaws in Workspace App for Windows, XenServer 8.4/9, and UniconOS, with risks ranging from remote code execution and DoS to VM-to-host compromise and privilege escalation. Fortinet patched 13 bugs across ten bulletins, highlighted by two critical issues — unauthenticated RCE in FortiSIEM (CVE-2026-106173) and an authentication bypass in the FortiSandbox GUI (CVE-2026-107656) — plus SQL injection flaws in FortiAnalyzer, FortiManager, and FortiMail. Ivanti's seven fixes span Connect Secure, Neurons Patch for Intune, and Xtraction; most notably, Connect Secure receives a fix for CVE-2026-104935, an actively exploited zero-day chain that let attackers gain root access on fully patched VPN appliances, while Neurons Patch for Intune gets a critical credential-leak fix (CVE-2026-109412) and Xtraction addresses a critical request-redirection flaw (CVE-2026-106042) that could expose EPMM connector credentials.

Microsoft's October 2026 security update sets a new record, resolving 1,413 CVEs in one release — more than the company fixed across all of 2025. The bulk of the fixes target Windows (1,102 flaws), with additional patches for Office (134), developer tools (61), SharePoint Server (36), plus Exchange Server, Azure and Defender. Among the fixed issues is an actively exploited zero-day, CVE-2026-96951 (CVSS 8.8), a Windows shell flaw that lets attackers bypass the Mark of the Web security check, allowing malicious downloaded files to run without triggering Defender SmartScreen or Office Protected View warnings. CISA has added it to its Known Exploited Vulnerabilities catalog and ordered federal agencies to remediate by October 16. Two other vulnerabilities were publicly known before patches shipped: CVE-2026-108398, a Windows Defender Application Control bypass considered unlikely to be exploited, and CVE-2026-92707, a denial-of-service bug in the MessagePack-CSharp component affecting Visual Studio and .NET, whose exploitability hasn't yet been assessed. Microsoft also introduced a new 'risk cohort' grouping system to help organizations prioritize which of this massive patch batch to apply first.

SAP's October 2026 Security Patch Day delivered 22 notes, four addressing critical-severity flaws across its product line. The most severe, CVE-2026-89111 (CVSS 10), is an OAuth2 SSRF bug in Cloud SDK for Python affecting multi-tenant BTP apps, letting an attacker manipulate the tenantsubdomain parameter to redirect token requests and steal client credentials. CVE-2026-89110 (CVSS 9.9) is a broken access control flaw in NetWeaver Visual Composer that lets a low-privileged user trigger internal operations on a connected remote system, while CVE-2026-89107 (CVSS 9.9) is a code injection issue in S/4HANA allowing authenticated users with a specific Fiori business role to run arbitrary ABAP code and seize the system. CVE-2026-89108 (CVSS 9.1) is a missing authorization check in Manufacturing Integration and Intelligence that lets an XMII Admin create a rogue administrator account on the Java application server. SAP also patched high-severity SQL injection, crash, and bypass issues in Commerce Cloud, NetWeaver, and ABAP Platform; no active exploitation has been reported, but customers should apply the fixes promptly.

Researchers at Doyensec managed to execute their own code inside a Windows VBS enclave, a memory region normally shielded even from an attacker who already controls the kernel or has SYSTEM privileges. Their target was SFAPE.dll, the enclave behind Microsoft Defender SmartScreen's Enhanced Phishing Protection feature, which stores a hashed copy of the user's login password to detect when it's typed into an unsafe app like Notepad or a phishing site. By calling the enclave's Initialize function twice, they triggered an unchecked four-byte write that corrupted an internal memory marker, then combined it with a timing flaw in another function to hijack the enclave's object lookup and swap out its list of permitted code — achieving arbitrary code execution inside the isolated world. The issue, tracked as CVE-2026-83498 with a CVSS score of 7.8, requires local access and was not seen exploited in the wild; it was patched by Microsoft on September 8, 2026 via update KB5124008 for Windows 11 24H2 and 25H2, which also blocks the vulnerable enclave from loading at all. Notably, a similarly flawed Microsoft Edge enclave from 2023 (CVE-2023-36880) remains loadable, and since Windows lacks a blocklist for vulnerable enclaves comparable to its driver blocklist, attackers could still abuse such older signed-but-flawed enclaves to hide malicious activity.

Microsoft's October Patch Tuesday set a new high-water mark, delivering fixes for 1,413 vulnerabilities across Windows, Office, .NET Framework, SQL Server, Developer Tools and SharePoint Server—surpassing last month's previous record of nearly 1,000. Among them is CVE-2026-96951, an 8.8-severity command injection flaw in Windows Shell that lets attackers bypass security protections and evade Mark of the Web defenses via malicious files or links; Microsoft confirms this one is already being abused in real-world attacks, prompting CISA to add it to its Known Exploited Vulnerabilities catalog with a fix deadline of October 16. Two other flaws stand out: CVE-2026-108398, a publicly disclosed .NET Framework issue that can bypass Windows Defender Application Control without user interaction, and CVE-2026-92707, a denial-of-service bug in MessagePack-CSharp. Microsoft also patched a critical 9.8-rated Raw Image Extension flaw (CVE-2026-21413) that can trigger remote code execution merely by generating a thumbnail preview. With over 4,000 CVEs fixed this year alone, Microsoft is rolling out a new 'risk cohort' grouping system to help organizations prioritize which of the flood of bugs truly need urgent attention.

Unit 42 has detailed a prolonged breach of a telecom managed-service provider carried out by the cluster tracked as UNC5820 (also called CL-STA-0954). The attackers first abused CVE-2024-47575 to compromise a FortiManager appliance, registering a rogue virtual FortiGate device and stealing an archive containing configurations, credentials, cryptographic material and inventory data for over 500 managed FortiGate firewalls. After cracking reversible credentials and establishing VPN access, they exploited a second flaw, CVE-2025-61624, a path-traversal bug in the FortiOS auto-script engine, to gain root and plant PEERSTITCH, a new ARM64 Linux implant that runs only in memory, hooks the miglogd and iked daemons via ptrace injection, and communicates covertly through legitimate process traffic rather than its own network connections, making it very hard to spot with standard tools. The actors also abused Apache's ACME modmd feature for a reverse-shell C2 channel and apparently impersonated FortiManager itself to pull configurations from other firewalls, underscoring how compromising one central management platform can expose an entire managed fleet, including client networks, well beyond the point where the original flaw is patched.

Microsoft's October update rollout addresses more than 1,400 vulnerabilities across Windows, Windows Shell, .NET Framework and Office, far exceeding September's already unusual batch of nearly 1,000 fixes. Among them, CVE-2026-96951, a security feature bypass in Windows Shell affecting Windows 10, 11 and some Server editions, is confirmed to be under active exploitation. Roughly 78 of the bugs are rated critical, meaning many allow remote code execution with minimal user interaction, while CVE-2026-108398, a .NET Framework security feature bypass, was publicly disclosed before patching though not yet seen exploited. Researchers note Microsoft has now surpassed 4,000 CVEs fixed this year, partly attributed to AI-assisted bug discovery, and recommend organizations prioritize the 26 flaws already listed in CISA's Known Exploited Vulnerabilities catalog.

Microsoft's October 2026 Patch Tuesday set a new record, fixing 1,413 vulnerabilities across its product line, including 80 rated critical and two zero-day flaws. One of the zero-days, CVE-2026-96951 (CVSS 8.8) in Windows Shell, is already being exploited in the wild; it lets attackers bypass the Mark of the Web tagging system, defeating downstream protections like SmartScreen, Office Protected View, macro blocking, and Smart App Control without triggering any warning to the user. The second zero-day, CVE-2026-108398 (CVSS 8.8) in .NET Framework, requires local low-privilege access and is considered less likely to be exploited, though researchers warn it's a potent post-compromise tool for bypassing Windows Defender Application Control. Beyond Windows, which accounted for the bulk of the fixes, analysts flagged a critical zero-click Outlook Reading Pane RCE (CVE-2026-70242, CVSS 9.8), a SharePoint privilege-escalation bug (CVE-2026-102866), and a Microsoft QUIC buffer overflow (CVE-2026-105016, CVSS 9.8) as top priorities for immediate patching.

Wordfence researchers used their AI-driven Argus tool to uncover CVE-2026-39919, a critical heap buffer overflow in Ghostscript, the engine behind PDF rendering and thumbnail generation often invoked via PHP's Imagick extension and ImageMagick. By embedding a malformed JPEG 2000 image inside a PDF, attackers can force Ghostscript to write 1,024 bytes into a 256-byte buffer, corrupting memory and potentially achieving unauthenticated remote code execution. The team confirmed the vulnerable code exists across Ghostscript versions 9.21 through 10.07.1, impacting ten tested Linux distributions (Debian, Ubuntu, Fedora, Alpine, AlmaLinux) and the official WordPress Docker image, though real-world exploitation proved inconsistent and depends on factors like PDF fonts, language settings, and system libraries. On WordPress sites specifically, abuse requires PDF upload capability, automatic thumbnail generation, and a Ghostscript-based processing chain — conditions not met by default installs lacking anonymous uploads. Artifex patched the flaw in Ghostscript 10.08.0 on September 8, with backports available for older branches; admins should upgrade via their hosting provider, rebuild Docker images, or temporarily block PDF uploads and disable ImageMagick's PDF rendering if patching isn't immediately possible. Wordfence has not detected any active exploitation of this flaw so far.

Microsoft's October 2026 security release addresses a total of 1,415 vulnerabilities across its product line, 85 of which carry a "critical" severity rating. Cisco Talos flags that one flaw, CVE-2026-96951, a command-injection security feature bypass in Windows Shell with a CVSS score of 8.8, is already being actively exploited. Of the critical issues, 67 are remote code execution bugs, and Microsoft also lists several other high-impact vulnerabilities it believes are more likely to be exploited soon, including CVE-2026-70229 in Active Directory Domain Services and CVE-2026-70571 in Microsoft Exchange Server (both CVSS 8.1). Administrators are urged to prioritize patching given the scale of this month's update and the confirmed in-the-wild exploitation.

Microsoft's October 2026 Patch Tuesday delivers fixes for a record 1,411 vulnerabilities, including 70 rated Critical (57 remote code execution, 11 elevation of privilege, 2 spoofing) and three zero-day issues. One zero-day, a command-injection flaw in Windows Shell that lets attackers bypass the Mark of the Web security feature, is confirmed to be actively exploited in the wild, though Microsoft has not disclosed attack details or who reported it. The other two zero-days are publicly disclosed but not yet exploited: a .NET Framework security-feature-bypass bug (found by Google Project Zero's James Forshaw) that can defeat Windows Defender Application Control and let a specially crafted assembly run inside a protected process without user interaction, and a denial-of-service flaw in MessagePack-CSharp tied to uncontrolled memory allocation in .NET. The release coincides with Microsoft ending support for Windows 11 24H2 Home/Pro and Windows 10 Enterprise LTSB 2016, making this their final security update, while Windows 11 23H2 Enterprise/Education/IoT editions lose updates starting next month.

Microsoft's October 2026 Patch Tuesday addresses a record 1,417 vulnerabilities, 76 of them rated critical, while 7 critical issues require no action from administrators. One flaw was publicly known before release and one is already being actively exploited. CVE-2026-96951, a Windows Shell security feature bypass, lets attackers sidestep the "Mark of the Web" protection, making it easier to run malicious code once a victim opens a file sent via email or instant messaging. CVE-2026-108398, a .NET framework security feature bypass that could let attackers evade application control, had public details before the patch but no observed exploitation. The update batch also covers 10 remote code execution bugs in the Windows DHCP Server (five critical), a high-complexity Exchange Server RCE tracked as CVE-2026-70571 triggered during email content conversion, and an unusually large set of Office fixes that administrators should prioritize deploying quickly.

Ivanti has disclosed active exploitation of a high-severity zero-day, CVE-2026-104935, affecting its Connect Secure (ICS) VPN appliances, allowing unauthenticated remote attackers to run arbitrary commands by bypassing authentication. Security firm Volexity uncovered the attacks while responding to a customer incident, finding that a newly identified group it calls UTA0576 chained three separate zero-day flaws to obtain root-level, unauthenticated code execution on an appliance that was already fully patched. CVE-2026-104935 accounts for two of the three bugs — the authentication bypass and a remote code execution flaw — which Ivanti has now fixed in ICS releases 22.8R2.5, 25.1.1.2, 25.1.3.0, and 25.1.3.1, while a third privilege-escalation issue is still awaiting a patch expected in coming weeks. Volexity traced exploitation back to October 3 and observed the attackers stealing stored domain credentials, hunting for LAPS passwords and BitLocker keys in Active Directory, running password-spraying and Kerberos attacks, and tampering with logs, though no web shell or backdoor was found; the firm suspects the group — possibly aided by LLM tools — had criminal rather than espionage goals and was likely preparing for broader lateral movement before being stopped.

Docker has closed a critical Docker Hub weakness, tracked as CVE-2026-105445 with a CVSS 4.0 score of 9.4, that allowed one user's personal access token to act as a different account when converting a personal profile into an organization. Researchers Peter Molnar and Krisztian Hozak of mptrdev.com, who dubbed the bug GhostPat, found that naming someone else as the new organization owner silently moved that person's tokens over without their knowledge or confirmation, letting an attacker push, pull or delete images under the victim's identity. The issue existed from August 26, 2024 until Docker shipped a server-side fix on September 30, 2026 — no customer action or update is required to apply it. Docker says a review of roughly 304,000 pushes made with transferred tokens found no sign of actual abuse, but it still revoked affected tokens on October 8 and emailed impacted users on October 2, urging everyone to audit their personal access tokens and switch automation to organization-level tokens instead.

Volexity disclosed that a threat actor it tracks as UTA0576 strung together three previously unknown vulnerabilities in Ivanti's Connect Secure VPN appliance to gain unauthenticated root access, bypassing the device's Integrity Checker Tool by patching library code only in memory. The first two bugs — an authentication bypass combined with a remote-code-execution issue, tracked jointly as CVE-2026-104935 and rated 8.6 by Ivanti — let the attacker execute commands as the cgi-server user on a fully patched appliance running version 22.8R2.4; a third, unpatched privilege-escalation flaw was then used to reach root. Ivanti has fixed the first two issues in version 25.1.3.0 and above, confirms active in-the-wild exploitation, and says a fix for the escalation bug is still weeks away, while CISA labels the exploitation status as active with total technical impact. The attacker stole LDAP credentials, scraped Active Directory for passwords, and password-sprayed accounts without dropping any backdoor, re-running the full exploit chain for every command; the same advisory also patches unrelated, unexploited critical flaws (CVSS 9.9) in Ivanti Xtraction and Neurons Patch for Intune.

SAP has published an advisory for a maximum-severity flaw, tracked as CVE-2026-89111 with a CVSS score of 10.0, in its Cloud SDK for Python, the official client library used by multi-tenant applications on SAP Business Technology Platform. The bug lets an attacker who controls a tenant's subdomain trick the SDK into sending the application's OAuth2 client ID and secret to a server they operate, a classic server-side request forgery issue, because the library failed to validate that the subdomain was a simple alphanumeric name before building the token-request URL. SAP says version 0.45.2, shipped on August 21, fixes the flaw, but research by The Hacker News found that six subsequent releases (0.53.0 through 0.55.0) reintroduced the missing check for calls through the Destination service module, and that two 0.36-series builds released after the fix never had the protection at all; the real fix for those paths only landed with 0.55.1. The advisory currently lists no version beyond 0.45.2 as vulnerable and gives no guidance on whether affected credentials should be rotated, so administrators running any release outside 0.45.2–0.52.2 or 0.55.1–0.58.4 should upgrade immediately or validate tenant subdomains against a known allow-list before passing them to the SDK.

Ivanti has released an urgent patch for a high-severity authentication bypass flaw in its Connect Secure (ICS) remote access gateway, warning that attackers are already exploiting it. The vulnerability, cataloged as CVE-2026-104935 and discovered by Volexity, affects ICS versions 22.8R2.4 and earlier, 25.1.2.2 and earlier, and 25.1.1.1 and earlier. It allows unauthenticated remote attackers to run arbitrary commands on vulnerable appliances without any user interaction, making it an easy entry point into corporate networks. Ivanti confirmed limited real-world exploitation and is advising customers to upgrade immediately and routinely run its Integrity Checker Tool, while Shadowserver reports more than 5,500 ICS instances still reachable from the internet. This marks yet another in a long string of Connect Secure zero-days abused by attackers in recent years, following flaws like CVE-2023-46805, CVE-2024-21887, CVE-2024-21893, CVE-2025-0282 and CVE-2025-22457.

Docker has patched a critical vulnerability, tracked as CVE-2026-105445, in Docker Hub that allowed attackers to hijack another user's personal access token (PAT) by converting their own account into an organization and designating the victim as its owner. Because the reassignment process didn't rotate the token's secret or require the target's consent, the attacker could then authenticate as that user within the PAT's scope and push, pull, or delete images in their repositories and connected organizations — a serious supply-chain risk since it could enable publishing malicious images under a trusted identity. The bug, present since August 26, 2024, was reported on September 30, 2026, and fixed in production within four hours; Docker says a review of 304,000 token-based pushes found no sign of abuse. Affected accounts were notified on October 2 and had their transferred tokens deactivated on October 8, while Docker Hub now requires explicit consent before naming someone an organization owner and no longer auto-transfers PATs during conversions. Docker clarified the flaw did not grant Docker Hub web console access or let attackers alter passwords, email addresses, or two-factor authentication settings, so account takeover in the traditional sense wasn't possible. Users who never received a notification from Docker are considered unaffected and don't need to act.

Check Point Research examined LiteLLM, a popular open-source gateway for routing requests to LLM providers, and found serious flaws in version 1.77.7, now fixed in 1.83.9. The core issue, tracked as CVE-2026-35029, was a missing authorization check on the /config/update endpoint, letting any logged-in user—even one with minimal privileges—rewrite the proxy's live configuration. From there, attackers could inject environment variables to read arbitrary files, overwrite admin login credentials to seize full control of the gateway, or register a malicious Python module from cloud storage to achieve remote code execution. A second flaw, CVE-2026-59823, resided in the "Clientside LLM Credentials" feature and let attackers abuse the userconfig parameter to redirect outgoing API calls via SSRF, capturing the provider's real API keys on an attacker-run server. Researchers also flagged weakly hashed, unsalted passwords and exposed password hashes as additional risks, underscoring that even fast-moving AI infrastructure projects still need solid access-control and input-validation fundamentals.