Cyber Risk Barometer
Home › CERT round-up

Global CERT round-up: September 2026

Two Citrix zero-days exploited on four continents, Iranian spyware exposed, Denmark's threat level raised to "high", and the EU's first product-reporting duties. What national CERTs and vendor research teams warned about in September.

Cyber Risk Barometer · 15 October 2026

Every month the Cyber Risk Barometer reads the public feeds of more than 120 national computer emergency response teams (CERTs) and 21 vendor research teams, and sets out what they were warning about. This is the first round-up. Every item links to the original advisory.

The month in brief

September ended with a scramble over two zero-day flaws in Citrix NetScaler ADC and Gateway, the remote-access boxes many organisations use as their front door. National teams on four continents warned that they were being exploited. Earlier in the month, actively exploited flaws in F5 BIG-IP APM, WordPress, Cisco Identity Services Engine and SonicWall's SMA1000 remote-access gateways set off similar waves of alerts, and Cisco later confirmed that an authentication bypass in its Catalyst SD-WAN Manager had been under attack during September. State activity was a steady theme: the UK and allies exposed Iranian spyware aimed at dissidents and journalists, Denmark raised its threat level for destructive cyber attacks to "high", and researchers tracked Russian, Chinese and Iran-linked campaigns. Attackers' use of AI also moved from theory to practice: Canada reported AI being used in a global campaign against industrial controllers, and Cisco Talos described what it calls the first reported autonomous AI command-and-control implant. On regulation, the EU Cyber Resilience Act's reporting duties for product makers took effect, Romania issued its first fine for failing to report an incident, and the UK's Information Commission served the Metropolitan Police with a reprimand and an enforcement notice.

Key stories

  • Citrix NetScaler zero-days exploited worldwide. Two flaws (CVE-2026-88771 and CVE-2026-88772) in widely used remote-access appliances were exploited before patches were out, which puts any unpatched organisation at risk of a break-in. Flagged by NCSC (UK), CERT-FR (France), CERT-SE (Sweden), CSIRT-IE (Ireland), CSIRT-IT (Italy), NCSC-NL (Netherlands), DNSC (Romania), SI-CERT (Slovenia), SingCERT (Singapore), GovCERT.HK and HKCERT (Hong Kong), KrCERT/CC (South Korea), CERT-QC (Canada) and CIRT/cm (Cameroon), with threat briefs from Palo Alto Unit 42 and Google Mandiant. NCSC UK · Unit 42
  • UK and allies expose Iranian spyware aimed at dissidents and journalists. A joint public attribution of Iranian state hacking against activists, dissidents and journalists. NCSC (UK). NCSC UK
  • Denmark raises the threat of destructive cyber attacks to "high". A national agency formally judging that attacks meant to destroy or disrupt, not just steal, are now more likely. CFCS (Denmark). CFCS
  • F5 BIG-IP APM flaw actively exploited. A critical bug in another widely used remote-access product was under attack, flagged by several European and Asian teams. NCSC-NL (Netherlands), DNSC (Romania), CERT-SE (Sweden), CSIRT-IE (Ireland), HKCERT (Hong Kong). DNSC
  • AI used in a global campaign to disrupt industrial controllers. Attackers are using AI to find and disrupt internet-exposed programmable logic controllers (the small computers that run machinery and utilities). CCCS (Canada). CCCS
  • WordPress core flaw actively exploited. A flaw in the software behind a large share of the world's websites was being used in attacks. NCSC-NL (Netherlands), SingCERT (Singapore), DNSC (Romania), HKCERT (Hong Kong). SingCERT
  • Cisco Identity Services Engine flaw actively exploited. A critical bug in Cisco's network access control product was under attack; CERT-SE separately warned of an exploited flaw in Cisco Secure Email Gateway. CSIRT.SK (Slovakia), SingCERT (Singapore), DNSC (Romania), CSIRT-IE (Ireland), CERT-SE (Sweden). CSIRT.SK · CERT-SE
  • Cisco Catalyst SD-WAN Manager authentication bypass exploited. A critical flaw (CVE-2026-76504, CVSS 9.8) lets an unauthenticated attacker send a crafted request to the management API and gain admin access to the system that controls an organisation's wide-area network. Cisco says it became aware of active exploitation in September; there is no workaround, only the update. Cisco PSIRT (US). Cisco PSIRT
  • SonicWall SMA1000 remote-access gateways under attack. CERT-FR raised an alert on 2 September for two flaws in the SMA1000 series (CVE-2026-83548, an unauthenticated server-side request forgery, and CVE-2026-83549, remote code execution by an admin) that SonicWall says are being actively exploited. Another remote-access product on the front line, alongside Citrix, F5 and Cisco. CERT-FR (France). CERT-FR
  • Metabase dashboards compromised through a SQL injection flaw. CERT-FR says it knows of many compromised Metabase installations (the open-source business-intelligence tool many teams expose to the internet) through CVE-2026-72898, patched in August; attempts show up in the application logs. CERT-FR (France). CERT-FR
  • ShinyHunters renews mass exploitation of Oracle PeopleSoft. The ShinyHunters group is back to hitting PeopleSoft (widely used for HR and payroll), with hidden web shells (back doors) being planted. Google Mandiant (US); also flagged by UZCERT (Uzbekistan). Mandiant
  • Several state-backed groups quickly adopt a Chrome and Windows zero-day chain ("BlueMoon"). Shows how fast a new exploit spreads between state hacking teams once it appears. Proofpoint (US). Proofpoint
  • First reported autonomous AI command-and-control implant. Talos describes malware that uses AI to run itself rather than waiting for a human operator. Cisco Talos (US). Talos
  • EU Cyber Resilience Act reporting duties take effect. Makers of digital products sold in the EU must now report vulnerabilities in their products. CCB (Belgium), CFCS (Denmark), with an early review from CERT.at (Austria). CCB · CERT.at
  • Romania issues its first fine for failing to report an incident. The first penalty under the country's cyber security emergency ordinance (OUG 155/2024) for breaching the duty to notify. DNSC (Romania). DNSC
  • UK regulator serves the Metropolitan Police with a reprimand and an enforcement notice. The Information Commission (the ICO) found the force had used third-party platforms without authorisation to share personal information about people suspected of shoplifting, breaching three sections of the Data Protection Act 2018. A rare combined action against a police force. Information Commission (UK). Enforcement notice · Reprimand
  • US agencies warn of Chinese AI firms "distilling" US AI models at industrial scale. CISA, NSA and FBI say China-based AI companies are copying US models' behaviour to shortcut their own development. CISA (US). CISA
  • Apple iPhone and iPad flaw used in real attacks. UZCERT says a high-severity flaw was exploited in live attacks; Hong Kong and Singapore issued high-threat alerts the same week. UZCERT (Uzbekistan), GovCERT.HK (Hong Kong), SingCERT (Singapore). UZCERT · GovCERT.HK
  • Albania analyses HEAVYGRAM malware linked to Iran's intelligence ministry (MOIS). A national technical write-up of a state campaign. NCSA (Albania). AKSK

Around the world

Europe. Citrix, F5, WordPress and Cisco alerts dominated, with the Netherlands also warning of actively exploited zero-days in the Zammad helpdesk software and Italy reporting exploitation of new MikroTik RouterOS router flaws (Slovenia and Croatia also issued critical MikroTik warnings). The Netherlands and Belgium both said AI is speeding up and enlarging the threat and urged basic security now. Recorded Future reported Russia escalating hybrid attacks across Europe, and CERT-UA (Ukraine) described group UAC-0277 using "ClickFix" lures (fake prompts that trick users into running commands) on hacked websites to spread the LUNEXSTEALER password stealer. Switzerland's Federal Council plans a new Cybersecurity Act, and Norway's NSM published advice on defending against denial-of-service attacks.

Americas. Canada's CCCS issued a statement on reported activity targeting Government of Canada websites, urged heightened vigilance amid global tensions, and co-published joint guidance on detecting Active Directory compromises. CISA ran its national Cyber Storm X exercise and published guidance with NIST on protecting federal cloud identity systems from token theft. In Brazil, Check Point described a Chinese-speaking actor abusing government websites for search-engine spam, and Mandiant tracked the financially motivated group BREEZE COMET.

Asia-Pacific. Cisco Talos reported a China-linked group (UAT-11587) targeting government and policy bodies across Asia with the Antino backdoor, and published a review of ransomware in Japan in the first half of 2026, including signs of the Qilin gang using AI. CrowdStrike reported an unknown actor using an AI-driven tool against South Korean finance. SingCERT warned of the CARBONATO botnet hijacking exposed Docker servers; TWCERT/CC (Taiwan) flagged public exploit code for a Microsoft Exchange Server flaw (CVE-2026-62911) and a 12-year-old high-risk PostgreSQL bug; KrCERT/CC warned of phishing emails impersonating public bodies.

Middle East and Africa. Kaspersky reported the Mirage Kitten group targeting aviation and fintech across the Middle East and Africa with new malware. Cameroon's CIRT/cm issued a run of critical alerts on exploited network and remote-access flaws, and Benin's bjCSIRT warned of critical flaws in Zimbra, GLPI, n8n and Palo Alto PAN-OS. Togo's CERT.tg warned of fake investment and gambling platforms.

Vendor research. Microsoft tracked an unauthenticated command-injection flaw on internet-facing mail servers (CVE-2026-73570), Star Blizzard's new "RedFlick" phishing technique, Storm-3168's AI-driven cloud attacks using stolen service accounts, and "EvilTokens" device-code phishing. SentinelLabs found TraderTraitor backdoors on a victim with no crypto links, and Kaspersky described PAYLOAD ransomware abusing Active Directory Group Policy. Mandiant and Google's threat team published reviews of AI's effect on vulnerability discovery and adversaries' move "from prompting to autonomy". Cisco's September hardening release also fixed a 9.9-rated flaw in Secure Firewall Management Center (CVE-2026-20324) that lets a logged-in user run commands as root over the sftunnel link between devices; Cisco knows of no exploitation. Cisco PSIRT

How this round-up is made

The Barometer collects the public news and alert feeds of 126 national and government CERTs, 21 vendor research teams and ten news alerts every day, and keeps the last 60 days. Each month the items are grouped by region and read through; the stories above are the ones the most teams warned about, or that mark a change in how attackers or regulators behave. Nothing is taken from private or members-only sources, and no contact details from the FIRST directory are collected.

Nine teams could not be read this month, so their warnings are not included: ACSC (Australia), CIRT-BS (Bahamas), CERT.BY (Belarus), CARICERT (Curaçao), EG-CERT (Egypt), KE-CIRT/CC (Kenya), MyCERT (Malaysia), ThaiCERT (Thailand) and UG-CERT (Uganda). Their sites either block automated readers or have certificates that cannot be verified.

Updated 19 October 2026: a second pass over the month's feeds, including Cisco's critical advisories and the UK Information Commission's enforcement listing, added the Cisco SD-WAN Manager, SonicWall, Metabase and Metropolitan Police items, and mwCERT (Malawi) could be read after all.

Get it by email

The Cyber Risk Barometer twice a year, the monthly global CERT round-up and new articles. Nothing else.

We store only your email address and the date you signed up, use it only to send these emails, and never share it. Every email has an unsubscribe link.