Overview
Nexwall Firewall 26.2.1 adds Advanced Protection and Collective Protection (AI), with one page to control and watch them in Protection Services, and measures the active devices and the throughput for the license size. It keeps everything of 26.1.3: updates that work from System > Update, separate block and inspect choices in Web Protection, block pages, the content filter, files tested in the Nexwall sandbox, Threat Analytics and Traffic Analytics, the new interface (3.0, light and dark), and the installer ISO for virtual machines, USB sticks and appliances.
- Version
- 26.2.1
- Status
- Stable: image and installer ISO built and tested
- Platform
- x86-64, UEFI boot
- Base system
- OpenWrt 25.12.5, kernel 6.12
- Interface languages
- English, Spanish, Brazilian Portuguese
- Licensed features
- Trial included
What is new
The first five items are new in 26.2.1; the others come from 26.1.x.
NewAdvanced and Collective Protection page
Protection Services has a new page with one switch for each protection, the real state of both engines (running, degraded or stopped), a resource monitor, model and list updates with an Update now button, recent advisory events and the privacy notice.
NewAdvanced Protection: behaviour scoring
The nexwall-atp service scores how each internal host talks to the internet (flows and DNS) with a model trained by Nexwall. A high score raises an advisory event; it never blocks. Models are signed, verified with the license server key before use, and the previous one is kept for rollback. The service is limited to about 3% of one core. The first models were trained on laboratory traffic (beacons, scans, floods): they do not detect other behaviours and can flag normal periodic check-ins. Treat events as indications to review, not verdicts.
Firewalls report suspicious addresses and domains to the Nexwall server. An indicator reported by at least 3 different firewalls within 14 days is published automatically; one reported by fewer goes to Nexwall's review page and is published only when confirmed. Published indicators expire after 14 days unless seen again, and rejected ones are not published again. They are distributed as two threat-list categories (atp_ip and atp_dns) and blocked on firewalls with Collective Protection on. A report from a single firewall never blocks traffic on other firewalls by itself.
NewReport sharing is on by default
Sharing is on by default, explained in the setup wizard and on the page. Reports contain only the external address or domain, the kind of behaviour, a score and the date, signed by the firewall. Traffic content, files, user names, internal addresses and browsing history are never sent. Sharing can be switched off at any time; the firewall still receives the collective lists if Collective Protection is on.
NewUsage measurement for the license size
The firewall counts active devices (including OpenVPN remote-access users) and throughput on the internal networks, and sends only these numbers with the daily license check. IPsec and WireGuard tunnels are not measured. Nothing is blocked when a size is exceeded; the figures let Nexwall and the partner check the license size. No addresses leave the firewall.
NewBlock page for blocked websites
When a rule with "Show a block page" blocks a website category, the user sees a page in Portuguese, Spanish or English (browser language) with the category, the rule and a reference code, which is also logged. The page is offered for categories the web proxy can open (up to 500,000 domains in total, smallest lists first); larger lists such as adult content keep the plain block. If the service that serves the page is down, the request fails and is never let through.
NewContent filter: School, Private and Work
Pages passing through the web proxy are scored from 0 to 100 for adult content, violence, drugs, gambling, hate speech and self-harm. A page that reaches the profile threshold is replaced by a block page showing the subject and score, or only recorded, as configured. Three profiles are provided (School strict, Private moderate, Work light), selected by the Standard, School and Restricted choices of the setup wizard; the Content filter tab lets you change threshold, subjects and where pages are read. Pages about study, health or news, and pages offering help, score lower. The filter reads text, not images, and only pages the proxy can open (unencrypted sites and decrypted categories).
NewIntelligent Threat Block: files tested in a sandbox
A file that passes the antivirus can be sent to the Nexwall sandbox, where it is opened and observed. With the hold option the download waits for the verdict (15 to 300 seconds, configurable) and a file judged malicious is blocked with an explanatory page; known verdicts are answered immediately. Limits per firewall: 60 files per hour, 500 files and 500 MB per day, 20 MB per file, 2 at a time (half during a trial). Available while the license includes it; settings, status and verdicts are in Web Protection and Threat Analytics.
NewThreat Analytics and Traffic Analytics
Two analytics pages follow the Dashboard. Threat Analytics shows what was stopped (block, antivirus, sandbox and advanced layers), devices at risk, web categories, top sources and reasons, and an event table kept for 35 days, with an executive report in PDF that is free in every plan. Traffic Analytics shows total traffic, busiest hour, top application and device, traffic by hour, and ranked lists of devices, applications, remote hosts and protocols for today, 7 or 30 days. The Monitor no longer has the Daily traffic tab.
New design in light and dark mode, consistent across all pages. The top bar has the page path, search (Ctrl+K), language, help, the status center and the unit chip (health, serial number, version, uptime, license). Each page opens with its tabs and a one-line purpose. Policy pages have tighter tables; Firewall Rules and Zones & Policies have a summary column (rules by action, busiest zone pairs, filter tags). The Dashboard shows the protection state and traffic for today, 7 or 30 days. Rules are deleted from a menu to avoid accidents.
NewWeb Protection rules: block and inspect are separate choices
In the rule editor each category can be blocked, allowed or inspected independently. Blocking is decided from the server name without opening the connection; inspecting opens a category in the web proxy for the antivirus, content filter and certificate checks (allowed categories) or the block page (blocked categories). Each column has an "inspect all" switch. A category that is blocked and not inspected costs the proxy nothing.
NewA QUIC switch in Web Protection
QUIC (HTTP/3) carries web traffic over UDP, where the proxy cannot read it. The new "Block QUIC (HTTP/3)" switch in Web Protection settings (on by default) refuses UDP port 443 for clients of web-filtering rules, so browsers fall back to HTTPS over TCP and block pages, SafeSearch, antivirus and content filter apply to all traffic. The refusal reaches the client immediately.
The Application Control, Network Protection (IPS), Web Protection and antivirus pages show a badge that is green only when the service is actually running (process up and bound to what it needs) and red when it is switched on but not running. The Dashboard uses the same reading.
NewNexwall VPN clients on the Remote Access page
The Remote Access (OpenVPN) page lists the latest Nexwall VPN clients: Windows (64-bit Intel/AMD, ARM64, 32-bit), Android, and Debian and Ubuntu packages (64-bit Intel/AMD and ARM64). The macOS client is not published yet. The clients are test builds and are not code-signed.
NewPacket loss and longer windows on WAN uplinks and Performance
Charts show packet loss instead of delivery rate. The WAN uplinks and Performance pages offer 1 hour (default), 6, 12 and 24 hours, 7 and 30 days, with current, peak and average values above each chart. For long windows each point is the highest value of its interval (lowest for free memory), so short spikes remain visible. History is kept for 7 days, or a year with a data disk.
NewLog Viewer: Web Protection tab
A new tab lists all web traffic handled by the proxy, allowed or blocked, including block pages, files stopped by the antivirus or sandbox, content blocks and refused requests, plus service messages. Intrusion prevention alerts and proxy requests are now sent to the Log Viewer.
The release is also provided as an installer ISO. Write it to a USB stick (for example with Rufus) and boot the appliance in UEFI mode. The installer verifies its image, lists only internal disks, requires typing INSTALL before erasing a disk, and reads the written image back to compare it.
Improvements and changes
ImprovedWeb proxy fixes verified again
The fixes for sites whose DNS answers change (the proxy uses the address the client chose, no status 409) and for ping and ICMP errors (no proxy timeouts) were verified again on a firewall running the previous image: 6 of 6 requests with a mismatched destination returned 200 and no 409 in 23,100 log lines; 160 of 160 proxied requests succeeded with and without two ping streams (p95 13.7 ms and 14.4 ms, 400 of 400 echo replies each); ICMP port unreachable reached the client in 0.05 s.
ImprovedProtection overview shows the new engines
The protection layers in the interface include Advanced Protection and Collective Protection with their real state (running, degraded, stopped, off or not in the plan).
ImprovedSystem > Update installs firewall fixes
Check for updates now finds the firewall packages, no longer fails the first time, and installs them. The update server index was incomplete and the image trusted a different key from the one that signs updates; both are fixed.
The Dashboard shows the number of rules loaded for the chosen policy, the same as the IPS Settings tab. It previously included catalog rules that only raise alerts.
ImprovedTop talkers on busy networks
With many connections the statistics window could be read as one second or less, showing rates several times too high or an empty list. The window is now at least ten seconds and measured to the millisecond.
ImprovedFewer "Unknown" in reports
Traffic to sites missing from the application catalog is listed under the site's domain instead of "Unknown". The catalog now includes a Nexwall application, so the firewall's own service traffic is classified.
ImprovedThreat Analytics sandbox state
The Sandbox stage shows whether the sandbox is working, not enabled on this firewall or not in the plan. Advanced Threat Protection and Collective Defense are still marked as coming in this version (see NXW-2018).
ImprovedThe traffic engine no longer stops responding under load
Two causes of the traffic engine ceasing to respond were identified from core dumps of a laboratory firewall under load and are fixed: two threads waiting for each other inside two Nexwall plugins, and the capture thread ending when the kernel dropped a queued packet. A watchdog now checks the engine every minute; if it stops responding it takes the engine out of the path (traffic continues without analysis) and restarts it, with a cooldown and an hourly restart limit. By default, when the engine cannot keep up, traffic is let through, not blocked.
ImprovedSites with changing DNS answers open behind the web proxy
Some sites (streaming services and other CDNs) answer the client and the proxy with different addresses. The proxy used to refuse such intercepted connections with error 409. It now connects to the address the client chose, as it already did for plain HTTP, without caching.
ImprovedBlocked internet traffic in the Log Viewer
The WAN zone logs what it rejects (at most 10 lines per minute), so the Firewall traffic tab shows blocked traffic as well as allowed.
ImprovedRefusals reach the client immediately
The "port unreachable" answer to a rejected connection and the messages path MTU discovery needs were held back by the traffic engine queue. They now pass, so a browser whose QUIC is refused falls back to HTTPS in milliseconds instead of after a timeout.
ImprovedThe web proxy under load
The proxy could run out of open files (limit 1,024) and the antivirus connection stayed off for minutes after a burst of failures. The limit is now 16,384, up to 256 files can be scanned at once, and the scanner connection is retried after 10 seconds.
ImprovedWireGuard tunnels and internet links
A tunnel no longer pins the route to its peer to whichever internet link was the default when it came up. On a firewall with two internet links, a stale route could cut off the license, updates and threat lists.
ImprovedApplication Control: engine switch and real state
Application Control has an engine switch, Web Protection turns the engine on when it needs it, and the Dashboard and pages show the real state of Network Protection (IPS) and Application Control. Top talkers and Traffic Analytics count the same flows.
ImprovedSafe by default for the web
An unconfigured firewall starts with the Standard profile applied to the web: the profile's website categories, the Work content filter and, when the license and storage allow, the antivirus. The setup wizard can change every choice.
ImprovedWAN at first start
A firewall started without the setup wizard has its WAN on the second network card (eth1), by DHCP, in the wan zone. On virtual machines the card was not yet visible when the default was decided; a later step in the same start now fixes this for unconfigured units.
ImprovedYouTube and SafeSearch with Application Control
The addresses used by SafeSearch and YouTube Restricted Mode are kept out of the analysis engine and the inspection, and DNS answers them with plain addresses, so Google, Bing and YouTube open in restricted mode even when Application Control blocks streaming or social networks for the same devices.
ImprovedLogs page uses the full window
The system log uses the full width and height, with search, row count and options in one row above it.