Fortinet Security Hardening Release: FortiOS 7.6.8, Who Is Affected and What the Upgrade Looks Like

The announcement in one paragraph

On 21 October Fortinet’s CISO team published a post announcing a Security Hardening Release across several products, with the fixed builds shipping on 22 October. The advisories themselves are not out yet. Fortinet says they will follow on Patch Tuesday, 10 November. The post says the issues came out of internal and external testing that included AI-assisted research, and it names Anthropic, OpenAI, Google and SpaceX among the organisations involved. It does not claim any of this is being exploited. What it does ask is that internet-facing devices go first.

So this is a patch notice with no CVE list attached. I would rather write about that honestly than invent detail, so everything below is either from Fortinet’s post, from the product documentation, or from my own lab.

Affected and fixed versions

ProductAffectedFixed in
FortiOS7.6.0 to 7.6.77.6.8, 8.0.2, 7.4.13, 7.2.14
FortiManager7.6.0 to 7.6.77.6.8, 8.0.2, 7.4.12, 7.2.13
FortiAnalyzer7.6.0 to 7.6.77.6.8, 8.0.2, 7.4.12, 7.2.13
FortiMailsee Fortinet’s post8.0.2, 7.6.7, 7.4.9
FortiAuthenticatorsee Fortinet’s post8.0.4, 6.6.11

Two things in that table are easy to trip over.

First, the numbering does not line up. FortiOS on the 7.4 train is fixed in 7.4.13, but FortiManager and FortiAnalyzer on the same train are fixed in 7.4.12. The same offset appears on 7.2 (7.2.14 against 7.2.13). If you track “the fixed 7.4 build” as one number across the estate, you will mark something as done when it is one release short.

Second, FortiOS 7.0 does not appear in the fixed list. If you still have 7.0 units in service, the answer is to plan a move to a supported train, not to wait for a 7.0 patch.

Patching before you can read what is fixed

I have written variations of this post for other vendors this year, and the pattern is the same each time: a hardening release with AI-assisted discovery named in the source line, fixes grouped together, and not much to hang a risk decision on. The Cisco versions are in the Catalyst SD-WAN hardening release, the IOS XE one, the IOS XR bundle and October’s round, and Arista said something similar about its own research.

What is different here is the order of events. Cisco’s advisories arrived with the fix. Fortinet is asking you to deploy first and read later. That is defensible, because a patch that sits out for nearly three weeks before the details land gives an attacker less to diff against, but it moves the decision onto you. You are choosing to upgrade production firewalls on the strength of a vendor blog post.

If your change process needs a CVE number to start, it will not start before 10 November. Decide now whether that is acceptable for the devices that face the internet.

Which devices go first

Fortinet’s guidance is internet-facing first. For a FortiGate, that is a short list to check, and none of it needs the new firmware:

show full-configuration system interface wan1 | grep allowaccess
show full-configuration system global | grep admin-sport
show full-configuration firewall local-in-policy
show full-configuration system admin | grep trusthost

You are looking for administrative services (HTTPS, SSH) allowed on a WAN interface, a local-in policy that does not end in an explicit deny, administrator accounts with no trusted hosts, and any service listening on an address you did not mean to publish. FortiOS grep takes one plain pattern, so run one line per interface.

I ran this on my own internet-facing 70G before upgrading it. I will leave the results out of a public post, but the exercise is worth doing because it tells you how urgent the upgrade really is. A unit with no admin access on its WAN, a deny-all local-in policy and trusted hosts on every admin is in a very different position from one with HTTPS open to the world. It does not remove the need to patch. It tells you the order.

Management planes come next. FortiManager and FortiAnalyzer are both on the affected list, and a FortiManager holds the keys to everything it manages. I covered what a compromised one leaves behind in Peerstitch, and the Fortibleed update is a reminder that Fortinet has had a busy autumn.

Upgrade order, and what I did

The sequence I used on my lab follows the usual rule that the management plane goes before the devices it manages:

  1. FortiManager and FortiAnalyzer to 7.6.8 (both report build 3788).
  2. The FortiGate VMs to 7.6.8.
  3. The physical FortiGate 70G, last, because it is the lab’s route to the internet.

On the upgraded VMs the FortiOS build moved from 7.6.6 build 3652 to 7.6.8 build 3744. Serial number and licence carried over, Firmware Signature still read certified, and a VM that already had private data encryption enabled still reported Private Encryption: Enable afterwards. The reboot reason changed from power cycle to warm reboot, which is a useful line to check afterwards because it tells you the unit went down through the upgrade path and not because someone pulled the power.

The 70G through FortiManager

The 70G was on 7.6.7 build 3704. The steps were:

  1. Retrieve the config in FortiManager first, so there is a fresh revision to go back to.
  2. Make sure the 7.6.8 image for the 70G is in FortiGuard, Firmware Images. Mine was already imported from a file I had checksummed against Fortinet’s published value.
  3. On the device, open Firmware Upgrade, choose Custom, pick FortiGate 7.6.8 build 3744 from the local images, and confirm the prompt about a customised image.

Progress went to 63% when the image finished sending, 80% when the unit reported the image upgrade done and rebooted, and then “Upgrade done successfully”. From start to finish it took under five minutes. FortiManager showed the device as out of sync while it rebooted and settled by itself afterwards. Everything behind the 70G lost its route out for those minutes, which is the thing to warn people about before you start.

Afterwards, on the unit:

FGT-70G-HUB # get system status
Version: FortiGate-70G v7.6.8,build3744,260929 (GA.M)
Firmware Signature: certified
Serial-Number: FGT70G<REDACTED>
Private Encryption: Disable
Branch point: 3744
Release Version Information: GA
Last reboot reason: warm reboot

And the part I like most, the rollback path:

FGT-70G-HUB # diagnose sys flash list
Partition  Image                                     Active
1          FGT70G-7.06-FW-build3744-260929           Yes
2          FGT70G-7.06-FW-build3704-260601           No

The previous release is sitting in the other partition. If 7.6.8 misbehaves on your hardware, that is the way back: boot the alternate partition and restore the config revision you took first. A FortiGate VM has no such partition, so the equivalent is a hypervisor snapshot before you start, which I had.

The other post-upgrade checks were unremarkable, which is the point of them. diagnose hardware test tpm still passed, the default route was in place, execute ping 8.8.8.8 returned 5 of 5 at about 10 ms, and get system auto-update status showed the unit reaching FortiGuard. My hub has no IPsec tunnels or BGP configured, so the empty output from get vpn ipsec tunnel summary and get router info bgp summary told me nothing either way. If yours does, those are the next two checks I would run.

Keep encryption changes out of the upgrade

I kept private data encryption off on the 70G and did not turn it on during the upgrade. Fortinet has documented problems upgrading HA pairs with private data encryption enabled, and an encrypted backup cannot be restored under a different key, which makes a rollback more complicated. Upgrade first, confirm the unit is healthy, then treat encryption as its own change with its own backup.

Two things the upgrade showed up in my lab

Neither is a flaw in 7.6.8. Both are what happens when you upgrade the management plane first, and both are worth checking in your own estate.

FortiManager 7.6.8 pushed settings an older FortiGate could not take

With FortiManager on 7.6.8 and one FortiGate VM still on 7.6.6, an install to that VM failed. The preview showed why: FortiManager’s device database wanted to write this under the custom-deep-inspection SSL inspection profile, for both https and dot:

config firewall ssl-ssh-profile
    edit "custom-deep-inspection"
        config https
            set udp-not-quic allow
        end
        config dot
            set udp-not-quic allow
        end
    next
end

udp-not-quic does not exist on FortiOS 7.6.6, so the install stopped there. The newer FortiOS builds took the same change without complaint. The lesson is not “do not upgrade management first”, because that is still the right order. It is that you should expect a FortiManager one step ahead of its devices to generate settings the older ones reject, run an install preview on every device before you install, and treat a failed install on a not-yet-upgraded device as a reason to upgrade it, not to fight the config.

My FortiAnalyzer was patched, healthy, and received nothing

FortiAnalyzer went to 7.6.8 without drama, and the link between FortiManager and FortiAnalyzer was up. Yet FortiAnalyzer’s device list showed my FortiGates as down with no logs. The cause was simple: none of them had log fortianalyzer setting enabled. Upgrading a collector does not make anything send to it.

On each FortiGate the fix is a short block, and a serial number is worth setting so certificate verification works:

config log fortianalyzer setting
    set status enable
    set server "10.10.0.20"
    set serial "FAZ-<REDACTED>"
    set upload-option realtime
    set reliable enable
end

After applying it the devices turned Up in FortiAnalyzer within a few minutes. So add one line to your post-upgrade checks: confirm every device is still sending logs to FortiAnalyzer, because you may find out that some never were.

A checklist for the weekend

  • Find every FortiOS, FortiManager, FortiAnalyzer, FortiMail and FortiAuthenticator version you run and compare it against the table above, product by product.
  • Rank devices by exposure, internet-facing first, using the four commands above.
  • Take a config backup or revision, and a snapshot for VMs, before touching anything.
  • Upgrade management first, then the devices it manages.
  • Run an install preview on every device before installing, especially any still on the old release.
  • Keep the old image available: alternate partition on hardware, snapshot on VMs.
  • Afterwards, confirm every device is still sending logs to FortiAnalyzer.
  • Put 10 November in the calendar and read the advisories properly when they land, because that is when you will learn what you patched.

Sources: