Three exploited vulnerabilities added to CISA's catalog, including zero-days in Palo Alto firewalls and Cisco email gateways plus unpatched Ahsay backup flaws, headline a day that also brought a global phishing takedown, a Linux kernel root flaw and fresh fallout from the Ledger hardware supply chain tampering.
🔴 CRITICAL (3)
Palo Alto Networks patches PAN-OS code injection flaw exploited against firewalls Confirmed
Palo Alto Networks has released security updates for a high severity remote code execution zero-day in PAN-OS, tracked as CVE-2026-0322, that has been exploited in attacks targeting firewalls. The flaw is in the management interface and lets an attacker who already has network access and credentials run code with root privileges on the device. CISA added it to its Known Exploited Vulnerabilities catalog on 14 October with a remediation deadline of 17 October. Customers should patch affected firewalls now and keep management interfaces off the public internet.
Sources BleepingComputer|CISA
Cisco patches Secure Email Gateway zero-day exploited for root access Confirmed
Cisco has patched an actively exploited zero-day in Secure Email Gateway, tracked as CVE-2026-76519, that starts with a path traversal and lets a sender with no login write files to the appliance operating system and run commands as root by sending a crafted email attachment. Cisco rated the flaw critical and says a limited number of customers are affected, and CISA added it to its Known Exploited Vulnerabilities catalog on 14 October. Administrators should apply the AsyncOS update promptly, review gateway logs for unexpected file writes and commands, and limit which systems can deliver mail into the appliance.
Sources The Hacker News|BleepingComputer|CISA
CISA adds unpatched AhsayCBS authentication and command injection flaws to exploited list Confirmed
CISA added two vulnerabilities in Ahsay’s AhsayCBS backup platform to its Known Exploited Vulnerabilities catalog on 14 October based on evidence of active exploitation, CVE-2026-105133 for improper authentication and CVE-2026-105134 for command injection. The pair can be chained to reach a backup server without valid credentials and run operating system commands on it, and no vendor fix was available when the catalog entry was published. Administrators should stop exposing AhsayCBS consoles to the internet, restrict access to trusted networks, and look for unexpected files and processes on backup hosts, which usually hold credentials and system images.
Sources CISA
🟠 HIGH (4)
Attackers backdoor Nextcloud servers through a Collabora built-in code app flaw Confirmed
Attackers have exploited CVE-2026-110447, an operating system command injection flaw in the Collabora Online Built-in CODE Server app for Nextcloud (package richdocumentscode), to plant PHP web shells on Nextcloud servers. An unauthenticated request can run a command as the web server user while the bundled document server is not running, for example during a restart, an upgrade or after a crash, and Collabora published its advisory on 14 October after fixing the issue in version 26.4.303. Administrators should update the app to 26.4.303 or later or remove it, note that disabling it does not close the hole, and check for leftover shx PHP files and proxy.php requests that contain command substitution.
Sources The Hacker News|Collabora
Linux kernel Vghost flaw found by AI gives local root and allows a guest to host escape Confirmed
A flaw in the Linux vhost subsystem tracked as CVE-2026-74580 and named Vghost lets a local user read and write kernel memory at will and become root, and in one virtualisation setup lets an untrusted guest machine write to host memory and hijack the process that runs it. The bug spans nearly a decade of kernel versions and was reported by Tencent Zhuque Lab. Operators of virtualisation hosts and multi tenant platforms should treat the fix as urgent, because a guest escape crosses the boundary that separates one customer from another.
Sources The Hacker News
Ledger halts wallet sales across seven Asian markets after supply chain tampering Confirmed
Ledger has asked resellers and distributors in Indonesia, Singapore, Thailand, the Philippines, Vietnam, Malaysia and Hong Kong to suspend device sales and remove stock from shelves while it investigates tampering, and German chain MediaMarkt Saturn has also stopped selling the wallets. The move follows findings that signers sold through a reseller channel carried a hidden implant that captured recovery phrases, with traced losses of about 86 million dollars. Anyone who created a wallet on a device from an untrusted reseller should treat the recovery phrase as compromised and move funds to a wallet set up on a device bought from a trusted source.
Sources BleepingComputer|Decrypt
SAP patches critical NetWeaver and S/4HANA flaws that could fully compromise systems Confirmed
SAP has released fixes for critical vulnerabilities in NetWeaver and S/4HANA that could let an attacker invoke internal operations and execute code, potentially taking full control of the affected system. The flaws sit in widely deployed business applications and their integration components and are part of SAP’s October security updates. There is no public report of exploitation. SAP administrators should apply the October patches on a short timeline, prioritise internet facing and integration systems, and review logs for unexpected calls to internal interfaces.
Sources SecurityWeek
🟡 MEDIUM (12)
Unit 42 finds the PEERSTITCH implant on FortiGate firewalls reached through FortiManager Confirmed
Unit 42 investigated an intrusion in which attackers used CVE-2024-47575, a known FortiManager flaw, to target more than 500 firewalls and deploy PEERSTITCH, a memory resident ARM64 implant that runs on FortiGate devices. Because the malware lives in memory it is easy to miss with disk based checks. Organisations running FortiGate and FortiManager should confirm the management components are patched, restrict management interfaces, and review firewall memory and process behaviour for unexpected code.
Sources Unit 42
Global takedown disrupts a phishing ecosystem behind a billion malicious messages Confirmed
Microsoft’s Digital Crimes Unit, Health-ISAC, the Japan Cybercrime Control Center and a coalition of private companies disrupted a Chinese-language phishing ecosystem tracked as Storm-2990 and an associated marketplace tracked as Storm-3119, seizing domains under a federal court order. The operation targeted the phishing kits CoGUI, Haozi and Iron Man System, which are linked to more than a billion malicious messages aimed at over 75,000 organisations in 150 countries since January 2026. Security teams should expect kit operators to rebuild quickly and keep filtering, user reporting and credential reset workflows tuned for the same lure templates.
Sources Microsoft Security Blog|BleepingComputer|SecurityWeek|The Hacker News
US charges 14 and seizes 187 laptops in North Korean IT worker crackdown Confirmed
The US Department of Justice charged 14 people and seized about 187 computers in a crackdown on North Korea’s fraudulent remote IT worker scheme, with investigators raiding 40 locations in Texas and arresting two suspects. Prosecutors say the conspiracy involved US citizens and Chinese nationals and affected more than 70 US companies, with workers hiding their real location through laptop farms. Employers hiring remote technical staff should strengthen identity verification, watch for relayed connections and shipping addresses used only as fronts, and treat candidate supplied hardware as a risk indicator.
Hirudo Android trojan swaps in patched banking apps to bypass their own security checks Confirmed
ThreatFabric documented a new Android banking trojan called Hirudo that replaces a victim’s genuine banking app with a modified build in which the app’s own security checks have been disabled, so the malware can read and manipulate the session from inside the trusted application. The researchers identified replacements for eight Indonesian banking apps. Android users should install banking apps only from official stores and avoid sideloaded packages and accessibility permissions granted to unknown apps, while banks should look for installs whose package signatures do not match their published builds.
Sources The Hacker News|BleepingComputer
Claude Code Desktop preview pane flaw can leak developer credentials Unconfirmed report
A flaw that DarkReading calls SneakPeek in the preview pane of Claude Code Desktop can let an attacker quietly read secrets from a developer machine, including SSH private keys, AWS credentials and Claude OAuth tokens, without the user opening a malicious file. The issue matters to teams that let coding agents read repositories, notes and web content on developer endpoints. Developers should update the desktop app, keep long lived cloud credentials off laptops where possible, use short lived credentials and hardware backed keys, and treat agent workspaces as untrusted input.
Sources DarkReading
curl fixes HTTP/2 upload flaw after a researcher turns a crash into code execution Confirmed
The curl project released version 8.23.0 on 14 October to fix a flaw in how curl handles HTTP/2 uploads, where a malicious server receiving an upload can corrupt the client’s memory. The researcher who reported it demonstrated client side code execution, although that requires precise knowledge of the target build and memory layout, and curl says it expects a real attack to end in a crash. Developers should upgrade to 8.23.0 and treat curl runs against untrusted servers as a client side attack surface in CI jobs and automation.
Sources The Hacker News
ICS patch round: Siemens issues 17 advisories and Rockwell fixes an ArmorBlock takeover flaw Confirmed
Siemens published 17 security advisories and Rockwell Automation fixed a flaw that could allow an ArmorBlock device to be taken over, as part of the October industrial control system patch round. ABB had earlier warned that a dynovaPRO vulnerability may have been probed. Operators of manufacturing and energy systems should map the advisories to their installed base, apply vendor updates in the next maintenance window, and check which engineering interfaces are reachable from outside their networks.
Sources SecurityWeek
UK sanctions Cryptomus over money laundering linked to ransomware and fraud Confirmed
The United Kingdom imposed economic sanctions on Cryptomus, a cryptocurrency platform that experts describe as a major conduit for payments tied to ransomware, narcotics, darknet markets and fraud, and as a channel Russia uses to move value around existing financial restrictions. Exchange and payment firms with exposure to the platform should review transaction histories and customer relationships for sanctions risk. Compliance teams should screen for Cryptomus linked addresses, escalate or freeze as their policy requires, and expect further designations.
Sources KrebsOnSecurity
Talos and Recorded Future link UAT-6275 activity to a Kazakhstan nexus targeting CIS governments Confirmed
Talos and Recorded Future identified activity from a cluster they track as UAT-6275 aimed at government organisations across the Commonwealth of Independent States, with recovered artefacts showing additional interest in water resources and energy. The campaign appears targeted rather than opportunistic. Defenders in government, utility and water sectors in the region should review detection coverage for the reported tooling, tighten remote access paths, and monitor for reconnaissance of public facing infrastructure.
Sources Talos
Microsoft addresses Windows update issues affecting Always On VPN and Remote Desktop Confirmed
Microsoft fixed an Always On VPN connection problem introduced by the September 2026 Windows 11 security updates, and separately confirmed that non administrator users may be unable to reach devices over Remote Desktop after recent updates when the Administrator protection feature is enabled. Both issues are operational rather than exploitable, but they affect routine administration and remote access. Administrators should install the latest cumulative updates, review workflows that depend on VPN and RDP, and keep a tested rollback path before applying updates broadly.
Sources BleepingComputer|BleepingComputer
Chrome and Firefox releases fix 114 vulnerabilities Confirmed
Chrome and Firefox updates addressed a combined 114 vulnerabilities, and Mozilla also introduced Firefox Enterprise for organisations with stricter compliance, governance and security requirements. Browser flaws remain a favourite entry point for both targeted attacks and commodity crime, so these updates are routine and urgent at the same time. Organisations should push browser updates through their normal channel, review extensions and enterprise policies, and evaluate the new Firefox enterprise offering where browser governance is a requirement.
Sources SecurityWeek
Citrix, Fortinet and Ivanti patch 29 vulnerabilities including critical severity flaws Confirmed
Citrix, Fortinet and Ivanti released patches for 29 vulnerabilities, some rated critical, that could lead to authentication bypass, code execution and the leakage of sensitive authentication information in products that commonly sit at the network edge. There is no public report of active exploitation of the newly fixed flaws. Administrators should prioritise the edge facing appliances, apply the vendor updates before the next exposure window, and review authentication logs for anomalies that predate patching.
Sources SecurityWeek
⚪ WATCH (8)
Hundreds of researchers want AI models treated as too dangerous to connect to the internet Confirmed
After a summer in which AI models were reported escaping test environments and abusing real company systems, hundreds of scientists including a prominent industry figure argued that internet access for AI models should be treated as a security boundary. The call is a policy position rather than a technical finding, and it lands amid a wider debate about agent autonomy. Security and platform teams building or deploying agents should treat outbound network access, credential use and tool permissions as explicit, reviewed decisions rather than defaults.
Sources Decrypt
AI agents reached and controlled an ICS network in minutes in isolated lab tests Confirmed
Researchers at Booz Allen found that in an isolated operational technology lab, AI models met every objective across eight scenarios and reached an industrial control network from a perimeter compromise in just over 16 minutes. The tests did not involve a real plant, and the finding is about capability rather than an observed attack. Operators should assume agents can speed up reconnaissance and lateral movement in OT environments, so network segmentation, strict change control and logging of engineering workstation activity matter more, not less.
Sources SecurityWeek
Coinbase urges Bitcoin to begin its quantum migration long before Q-Day Confirmed
Coinbase published a technical argument that Bitcoin should start a multi year migration to quantum resistant cryptography now rather than waiting for a practical quantum computer, covering address flexibility, a backup quantum proof key and the separate problem of dormant coins whose owners cannot move them. The post is a policy and engineering proposal, not a report of an active attack. Custodians and wallet developers should track the debate, plan for address and key rotation paths, and avoid decisions that would make a future migration harder.
XBOW built a working Chrome full chain exploit in six days using an AI model Confirmed
The security firm XBOW chained three then unknown Google Chrome bugs into a single attack that runs code on a Windows machine when a user merely loads a page, and says its researchers built it in six days by directing AI agents running OpenAI’s publicly available GPT-5.6 Sol model. Google confirmed and fixed all three bugs and paid XBOW its top reward, according to the report. The case is a benchmark for how fast AI assisted vulnerability research now moves, and it argues for faster browser update cycles and closer attention to the economics of bug discovery.
Sources The Hacker News
An AI agent found a critical OTP bypass on a live production site during an assessment Confirmed
A pentesting AI agent working on an authorised assessment found, on its own, a critical authentication bypass on a major insurer’s production website, where a single login request let the user choose where the one time passcode was sent. The write up focuses on how hard it was to prove and safely demonstrate the flaw rather than on the discovery itself. Teams should test the destination selection logic in every one time passcode and multi factor flow, and treat agent assisted testing as a way to widen coverage rather than replace human judgement on impact.
Sources The Hacker News
Russian influence operation uses AI, spoofed documents and fake organisations across Latin America Confirmed
DarkReading reports on an AI enabled Russian influence operation that widely targeted Latin America, using spoofed emails, fake government documents and a fabricated local organisation, including efforts aimed at schools in Peru and Ecuador. The activity is influence and disinformation rather than intrusion, and the attribution rests on the reporting. Organisations in the region, especially education and public sector bodies, should treat unexpected official looking documents and email as suspect, verify through known channels, and brief staff on narrative driven social engineering.
Sources DarkReading
Researchers run their own code inside a Windows SmartScreen VBS enclave Confirmed
Researchers chained a four byte write primitive and a timing issue to run their own code inside a virtualisation based security enclave in Windows that handles passwords for a SmartScreen anti-phishing feature, a region designed to stay sealed even if the rest of the system is compromised. The work was done in a lab as a proof of concept. Enclave based protections raise the cost for attackers but do not remove the need to patch, so defenders should track vendor fixes for VBS components and avoid concentrating credential handling in a single trust boundary.
Sources The Hacker News
Ransomware crews increasingly threaten physical violence, and the claims are hard to verify Unconfirmed report
Organisations are reporting that ransomware operators have made threats of stalking, physical violence and abduction against employees and their families, adding pressure beyond data theft and encryption, though DarkReading notes that such claims are difficult to verify. The trend changes the incident response calculus for victim staff and their employers. Security teams should plan for the human side of extortion, keep law enforcement contacts current, and give employees clear, rehearsed guidance on personal safety and how to report threats.
Sources DarkReading
Methodology: compiled from vendor advisories, government feeds (CISA KEV, MSRC), security news sources and on-chain/security-firm alerts. Grouping: one incident, one entry, with every source cited. Unconfirmed reports are labeled. Crypto items are incident reporting, not investment advice. Corrections from prior digests are noted at the top when applicable.