Our weekly round-up of significant threats and actively exploited vulnerabilities. It is awareness content, not a monitored threat intelligence service: always check vendor advisories against your own environment.
This week at a glance
Websites were the story of the week. Two critical PHP flaws (PHP2Shell) were exploited before fixes were released, stolen UpdraftCentral keys were used to plant backdoors on WordPress sites, and a researcher showed how one attacker used AI agents to break into more than 100 company websites. Arista patched a third VeloCloud Orchestrator zero-day, CISA added a chained set of Plex Media Server flaws to its exploited list, and UK, US, and German agencies issued a joint warning about SafePay ransomware.
Actively exploited vulnerabilities
PHP (CVE-2026-109618, CVE-2026-109619)
Two critical flaws in PHP, nicknamed PHP2Shell and rated CVSS 9.2, let an attacker run code on a web server without logging in. Details leaked before the planned release, and Amazon reports exploitation attempts from 17 October, four days before fixes were available. The UK NCSC urged organisations to act immediately, and advised taking sites offline if they cannot be patched quickly. A further PHP flaw affecting only non-default session settings has mitigations but no fix yet. Fix: update to PHP 8.5.12, 8.4.27, 8.3.36, or 8.2.35 (or confirm your host has), and move sites on unsupported PHP 8.1 or older to a supported version.
Source: UK NCSC
Arista VeloCloud Orchestrator (CVE-2026-114625)
An SQL injection flaw (CVSS 9.9) lets any authenticated user, even one with a read-only role, gain operator rights on on-premises VeloCloud SD-WAN Orchestrators. Arista says it is being exploited, the third exploited Orchestrator zero-day in four months, and CISA added it to its catalogue on 23 October. Fix: upgrade to 5.2.3.18, 6.1.3.10, 6.4.2.12, or later, restrict the web interface to trusted admin networks, and check logs for the IP addresses Arista published.
Source: Arista Security Advisory 0199
Plex Media Server (CVE-2026-96652, CVE-2026-96654, CVE-2026-96655, CVE-2026-96656)
CISA added four Plex Media Server flaws to its Known Exploited Vulnerabilities catalogue on 20 October. Chained together as “Plex2Shell”, they let an attacker with no account run code on an internet-exposed server, and attackers began using them the day after a public proof of concept appeared, installing cryptominers and residential proxy software. Fix: update to Plex Media Server 1.43.3.10861 or later and check for unfamiliar processes.
Source: BleepingComputer
EG Secure Solutions SiteGuard Server Edition (CVE-2026-106931)
A command injection flaw in the statistics feature of this host-based web application firewall, enabled by default, gives root or SYSTEM access. The vendor and JPCERT/CC have confirmed attacks, and JPCERT/CC warned users in Japan, where the product is widely used. Fix: save the vendor’s diagnostic information first (patching can erase traces), then update to Ver8.00 Update5 (or Ver7.00 Update6 on FreeBSD).
Source: BleepingComputer
Siemens Software Center (CVE-2026-110936)
Attackers are using a genuine, signed copy of Siemens Software Center to load a malicious DLL and install a remote access trojan, after tricking users with ClickFix lures delivered through the Romulus platform. Siemens says none of its own systems or update channels were breached. Fix: make sure every copy has updated to V4.0.9.8 or later and check for Siemens’ published indicators.
Source: NHS England Digital
Notable threats and campaigns
SafePay ransomware: joint #StopRansomware advisory
CISA, the FBI, and the US Secret Service, with the UK’s National Crime Agency, the North West Regional Organised Crime Unit, and Germany’s LKA Rheinland-Pfalz, warned that SafePay has hit more than 600 organisations since August 2024, across North America, South America, Europe, and Australia, with dozens of German organisations on its leak site. Affiliates mostly get in with stolen or guessed VPN and remote access credentials, sometimes through known edge-device flaws, then delete shadow copies and reset NAS devices holding backups before encrypting.
Source: CISA advisory AA26-293A
Stolen UpdraftCentral keys used to backdoor WordPress sites
TeamUpdraft says its website was compromised on 17 and 18 October and that attackers probably reached the database holding UpdraftCentral connection keys, which let the service manage connected WordPress sites. The stolen keys were used from 18 October to install backdoor plugins (Patchstack counted 99 different names in about 18 hours) and, in some reports, hidden administrator accounts. UpdraftPlus 1.26.13 and 2.26.13 disconnect all UpdraftCentral connections, but updating does not remove a backdoor already installed.
Source: Patchstack
Microsoft links the ConnectWise Cloud Backup attacks to Storm-0940
Microsoft attributes last week’s ConnectWise Cloud Backup supply chain attack to Storm-0940, which it describes as a China-state aligned actor; China has denied the claim. Microsoft says the group used ARTEX, an open-source agentic AI penetration testing framework, to automate password spraying and then abused the trusted backup app’s permissions to reach more than 2,250 Microsoft 365 tenants, in some cases gaining privileged access in under three hours.
Source: Microsoft Threat Intelligence
Also reported this week
- APT28 Hookedge espionage: Symantec reports that the Russian GRU-linked group used its Hookedge backdoor, sent by spear-phishing, against a naval shipbuilder, a military aircraft engine maker, and a law enforcement agency in two European NATO countries (The Record).
- LiteLLM supply chain attack: a malicious LiteLLM 1.107.0, published to PyPI and Docker on 21 October, stole cloud, AI, and other credentials from any Python environment it was installed in (Socket).
- AI agents used to breach websites: researcher NorForge and Reuters report that one Chinese-speaking attacker used autonomous AI agents to break into websites linked to more than 100 companies, aimed at e-commerce sites in the US, Canada, Australia, and Japan (Reuters via RTÉ).
- Athena StealthRoute phishing service: Abnormal AI details a phishing-as-a-service marketplace using AI-built lures and link cloaking, seen in about 89,700 messages to 2,280 organisations in the US and EU (Abnormal AI).
- Operation Serengeti III: Interpol reports 1,462 arrests and more than $105 million recovered in a summer crackdown on fraud, business email compromise, and ransomware across 20 African countries (Help Net Security).
Other patches worth knowing
- Oracle: the October 2026 Critical Patch Update fixes about 1,200 CVEs, including 95 rated critical.
- Fortinet FortiSIEM (CVE-2026-106173): a public proof of concept was released on 20 October for this critical flaw, and Canada’s Cyber Centre says it is being exploited, although Fortinet has not confirmed this. Upgrade 7.5.0 to 7.5.1 or block TCP port 27918.
- Nextcloud built-in Collabora CODE (CVE-2026-110447): covered last week, and added to CISA’s catalogue on 23 October.
- Palo Alto Networks PAN-OS (CVE-2026-0257): Unit 42 says ransomware groups are now exploiting this GlobalProtect flaw, fixed in June, so check that it is patched.
Practical steps to consider
- Ask your web host or developer, in writing, which PHP version your sites run and whether it is patched, and check WordPress sites for unknown plugins and administrator accounts (ISO/IEC 27001 control A.8.8, management of technical vulnerabilities).
- Review the third-party apps and services that hold standing access to your Microsoft 365 tenant and websites, such as backup and remote management tools, and remove any you no longer need (controls A.5.19 and A.5.21, supplier and ICT supply chain security).
- Enforce MFA on every VPN and remote access login, and keep at least one backup offline, as SafePay targets both (a core Cyber Essentials requirement).
This round-up is for general awareness only and is not a monitored threat intelligence service. If you would like help reviewing your vulnerability management or incident readiness, contact us.
