In the space of eight days in May 2026, two of the most widely deployed network platforms in Europe gained an actively exploited, unauthenticated vulnerability in the part of the product that is meant to be reachable only by administrators or by trusted peers. On 6 May the US Cybersecurity and Infrastructure Security Agency (CISA) added CVE-2026-0300, in the PAN-OS User-ID Authentication Portal, to its Known Exploited Vulnerabilities (KEV) catalog. On 14 May it added CVE-2026-20182, a maximum-severity authentication bypass in Cisco Catalyst SD-WAN. Looking back from November, the two cases share one lesson that applies well beyond these vendors: when an interface built for administration or control answers on the internet, the next critical bug in it becomes your incident. For the broader method, start from our guide to external attack surface management.
Key Takeaways
- CVE-2026-0300 lets an unauthenticated attacker run code as root on PA-Series and VM-Series firewalls through the User-ID Authentication Portal. Palo Alto Networks observed limited exploitation against portals exposed to untrusted IPs or the internet.
- CVE-2026-20182 (CVSS 10.0) gives an unauthenticated attacker administrative access to Cisco Catalyst SD-WAN control components, regardless of configuration. Cisco saw limited exploitation in May 2026 and published no workaround.
- The pattern is not new. Nine Cisco SD-WAN vulnerabilities entered the KEV catalog in 2026, from February to September.
- The root cause is exposure, not one bug: management planes, captive portals and controller interfaces reachable from networks that should never reach them.
- The fix is boring and effective: allow-lists, a separate management network, VPN-only administration and an inventory that matches what the internet actually sees.
Case 1: PAN-OS and the User-ID Authentication Portal
Palo Alto Networks published the advisory for CVE-2026-0300 on 5 May 2026. It is a buffer overflow in the User-ID Authentication Portal, also known as Captive Portal, that an unauthenticated attacker can trigger with crafted packets to execute code with root privileges on PA-Series and VM-Series firewalls. The vendor rated it CVSS-BT 9.3 with the highest urgency, and 8.7 where access is limited to adjacent networks. Its exploitation status was "attacked": the company reported limited exploitation targeting Authentication Portals exposed to untrusted IP addresses or to the internet. CISA added it to the KEV catalog the next day.
The conditions for exposure are the important part. A firewall is vulnerable when the portal is enabled and a management interface profile with Response Pages is applied to a Layer 3 interface in a zone where untrusted traffic can arrive. In other words, a feature meant for users on the internal network was answering on an interface that faced outward. The vendor's mitigations until the fixes arrived were to restrict portal access to trusted zones, disable Response Pages on interfaces facing untrusted zones, disable the portal if not needed, or apply a Threat Prevention signature. Prisma Access, Cloud NGFW and Panorama were not affected. CISA's KEV entry records that patches were available from 13 May.
Case 2: Cisco Catalyst SD-WAN, again
On 25 February 2026 CISA issued Emergency Directive 26-03 after the exploitation of CVE-2026-20127, an authentication bypass in the SD-WAN Controller and Manager, chained with an older path traversal flaw, CVE-2022-20775. US federal agencies had to inventory their SD-WAN systems, collect forensic artefacts, patch within about two days, hunt for compromise and apply hardening guidance.
On 14 May Cisco published CVE-2026-20182, with a CVSS base score of 10.0. The flaw is in the peering authentication used when control connections are set up. An unauthenticated remote attacker can send crafted requests, log in as an internal high-privileged account and use NETCONF to change the configuration of the SD-WAN fabric. The advisory covers the Controller, Manager and Validator in every deployment type, including Cisco-managed cloud, regardless of configuration. Cisco reported limited exploitation in May 2026 and listed no workaround: the only fix was to upgrade. CISA added it to KEV the same day and pointed to the ED 26-03 instructions.
The KEV catalog shows how persistent the problem was. Nine Cisco SD-WAN entries were added in 2026:
| Date added to KEV | CVE | What it allows |
|---|---|---|
| 25 February | CVE-2026-20127, CVE-2022-20775 | Authentication bypass to admin; path traversal to root |
| 20 April | CVE-2026-20122, CVE-2026-20133, CVE-2026-20128 | File overwrite through the API; information disclosure; recoverable stored credentials |
| 14 May | CVE-2026-20182 | Unauthenticated authentication bypass to admin (CVSS 10.0) |
| 9 June | CVE-2026-20245 | Command execution as root with a crafted file |
| 15 June | CVE-2026-20262 | Arbitrary file creation or overwrite |
| 30 September | CVE-2026-76504 | Unauthenticated access with admin privileges via crafted URI encoding |
Why the management plane is the real root cause
Every product will have critical bugs. What decides whether a bug becomes your incident is who can reach the vulnerable code. An administration web interface, an SSH service, a controller that accepts peering connections, a captive portal: each was designed for a small, known population. When they answer on the internet, the population becomes everyone, including the scanners that start looking for a new flaw within hours of its publication.
The data from 2026 points the same way. The 2026 Verizon DBIR found vulnerability exploitation to be the most common initial access vector, at 31% of breaches overall and 47% in Europe, the Middle East and Africa. It also found that only 26% of KEV vulnerabilities had been fully remediated, with a median of 43 days to full resolution. In Italy, the ACN annual report for 2025 notes that the highest numbers of vulnerabilities in the at-risk devices and services CSIRT Italia reported to their owners were in network devices and in firewall and proxy solutions. Patching faster is necessary. Removing exposure is what keeps the next bug from mattering.
Finding your own exposed management interfaces
The hard part is not knowing that management interfaces should be private. It is knowing where yours are. The inventory you have and the one the internet sees often differ: a branch firewall configured years ago, a test VM-Series instance in the cloud, a supplier-managed SD-WAN appliance, an interface profile copied from one zone to another.
- Start from the outside. List your public IP ranges and cloud addresses, and every domain and subdomain you own. Then look at what answers: which ports are open, which services identify themselves, which pages ask for a password.
- Look for the usual suspects. Web administration on 443, 4443 or 8443; SSH on 22; SNMP; NETCONF on 830; vendor-specific controller ports; login pages with a vendor's logo or favicon.
- Separate what must be public from what must not. A remote-access VPN portal has to answer on the internet. The firewall's management interface, its captive portal and an SD-WAN controller's administration do not.
- Check the configuration that creates exposure. On PAN-OS, review which interfaces carry a management profile and whether Response Pages or the Authentication Portal are enabled on interfaces in untrusted zones.
- Compare with the CMDB. Every host that answers and is not in the inventory is a finding in itself.
Locking it down
- Allow-lists first. Administration should be reachable only from defined source addresses. For SD-WAN, CISA's hardening guidance recommends not exposing the infrastructure publicly and placing a firewall in front of it that only admits the known IPs of the SD-WAN components.
- A separate management network. Move management interfaces into an isolated internal management VLAN or VRF, so that an administrator must first authenticate to the internal network.
- VPN or jump host with phishing-resistant MFA for every administrative session.
- Turn off what you do not use on outward-facing interfaces: response pages, portals, HTTP management, legacy SNMP versions.
- Keep logs off the device. Attackers edit local logs; send them to a SIEM and alert on new administrator accounts and on successful root logins, which CISA's guidance says should be treated as a root compromise.
- Back up configurations and compare them. An unexpected change in a firewall or controller configuration is an incident signal.
- Patch, then check for compromise. If a vulnerable interface was exposed during an exploitation window, an upgrade does not remove what an attacker may have left. CISA's SD-WAN guidance goes as far as redeploying the control components from patched images when root compromise is found.
ED 26-03 explained for European companies
Emergency directives are binding only on US federal civilian agencies. They still matter in Europe for three reasons. First, they are a strong signal: CISA issues them when exploitation is real and broad. Second, the required actions (inventory, collect evidence, patch, hunt, harden) are a ready-made checklist you can reuse. Third, CISA itself encourages all other organisations to apply the guidance. For NIS2 entities, these actions map naturally to the measures on vulnerability handling, access control and incident handling in Article 21. The KEV catalog dates are also a practical prioritisation input; our article on the vulnerability management lifecycle shows where they fit.
Where Orizon fits
RECON External starts from your domain name and shows what an attacker sees: each host with its IP addresses, technologies, open ports, status and TLS version; where your sites ask for a password, with a screenshot of each page; and login and admin panels recognised by their favicon hash, even on unusual ports. Detected products are matched to known CVEs and weighed by CISA KEV, Exploit-DB, Metasploit and EPSS, and each finding is then tested from the scan's own isolated server and marked Proven, To verify or Closed, with the evidence. RECON confirms that a vulnerability exists; it does not exploit it. Scans can be scheduled daily, so a management page that appears on a Tuesday is not discovered by someone else first.
When you need to know what an attacker could actually do with an exposed interface, Fireline runs an autonomous penetration test on the external perimeter covered by your signed authorisation.
Checklist
- List every public IP and domain, then everything that answers on them.
- Mark every administration interface, captive portal and controller port that is reachable from the internet.
- For each, decide: allow-list, move behind VPN, or switch off.
- On PAN-OS, check interface management profiles and Response Pages on untrusted zones.
- On SD-WAN, restrict control connections to known component IPs and isolate management interfaces.
- Send device logs to a SIEM and alert on new admin accounts and root logins.
- Subscribe to vendor advisories and the KEV catalog, and decide in hours, not weeks.
- Re-scan from the outside after every change and keep the evidence.
The next critical bug in an edge device is a matter of when. Whether it becomes your incident depends on whether its interface answers on the internet. To find out what yours expose today, run RECON External on your domains and fix what is proven first.