Critical (9.8) Linux Kernel Flaw: How to Tell If Your Servers Are Exposed and What to Do

If your company has Linux servers —a web server, a database, a cloud virtual machine, a homemade firewall— this week you have a concrete task: check the kernel. The flaw CVE-2026-98432, in the Linux kernel's IPv4 TCP code, is rated 9.8 out of 10 (critical) and can be exploited remotely, without credentials. A fix exists, but downloading the package is not enough: you have to reboot.
What CVE-2026-98432 is
According to Red Hat and Ubuntu, it is a race condition in the kernel's IPv4 networking code: when retransmitted TCP connection requests (or socket migrations) are processed at the same time, the kernel can free a socket's options while they are still in use. It is a use-after-free, a type of memory error. A remote attacker could trigger it with precisely timed packets and cause anything from a server crash (denial of service) to memory corruption or remote code execution. Red Hat and Ubuntu agree on the 9.8 score (CVSS 3.1) and on rating it critical; both state it needs no authentication or user interaction. Red Hat adds an important nuance: reliable code execution requires precise control of memory, so it is not trivial.
According to Bitsight, the kernel team disclosed it on October 15 and the defect has existed since Linux 4.4, so it affects a great many servers. The same source reports that Anthropic said it turned the flaw into a zero-click remote root exploit against a test server; this is a vendor claim, and as of the sources' dates no active exploitation on the internet had been reported. That can change, which is why it is wise to patch without waiting.
How to know whether you are exposed
- Inventory your Linux servers: physical, virtual, containers on a host and Linux-based network devices (routers, firewalls, NAS).
- Check the running kernel with
uname -r. Note that the version number alone is not enough, because distributions apply the fix to their own version (backport). Compare with your distribution's advisory. - Look up the status in your vendor's advisory. Ubuntu publishes a per-release table. In its latest revision (November 18, when we fact-checked this note), the
linuxkernel showed as fixed in 22.04 (5.15.0-203.213), 24.04 (6.8.0-149.149) and 26.04 (7.0.0-43.43); 14.04 is not affected; and 16.04, 18.04 and 20.04 remained vulnerable, with the fix in testing (Ubuntu says it will arrive through the ESM archive, for Ubuntu Pro users). We confirmed that table through ubuntu.com's listing in a search engine, because the site did not respond at the time of the check: confirm it in the advisory. Red Hat points to its bulletin RHSB-2026-016. The status changes almost daily, so check the page's date. - Prioritize anything that accepts connections from the internet: web servers, VPN, mail, databases exposed by mistake.
What to do: patch and reboot
Bitsight is clear: the only complete fix is installing your vendor's patched kernel and rebooting. Afterward, confirm with uname -r that the machine actually booted the new kernel; downloading the package does not protect the server. If you cannot reboot today, schedule the nearest maintenance window.
If you cannot patch yet: a temporary mitigation
Red Hat and Ubuntu document a stopgap: drop incoming IPv4 TCP packets that carry options in the IP header (normal traffic, UDP and IPv6 keep working). Ubuntu gives the rule for UFW, iptables and nftables; for example, with iptables: sudo iptables -I INPUT 1 -p tcp -m u32 --u32 "0>>24&0x0f=6:15" -j DROP. Three warnings the sources repeat: it is temporary, it can affect legitimate traffic (test it first), and in containers or machines in other network namespaces it must be applied inside each one. Also, Red Hat states that disabling MPTCP does not remove the risk. Once you install the fixed kernel and reboot, remove the rule.
What this means for your business
Most small businesses have no security team, but they do have servers that someone installed years ago and that "just work." This kind of flaw is exactly what gets exploited on forgotten machines. Three useful questions for your next IT meeting: do we know how many Linux servers we have? Who updates them? When were they last rebooted? If any answer is "I don't know," that is where the risk is. If you have systems on unsupported versions, plan their migration: a platform's end of support and an unpatched kernel are the same problem seen from different angles.
For system administrators
Automate detection: patch-management tools or simple queries with apt list --upgradable / dnf check-update, plus a tool like needrestart or your distribution's reboot-required marker. If you use live patching (livepatch), confirm with your vendor that it covers this specific CVE before assuming you are protected. And record in your inventory which machines need a reboot so they are not left "half-patched." For other recent advisories with the same logic, see the critical Atlassian flaw and the critical PHP fixes.
How we approach it at SimCodec
At SimCodec we help companies put their infrastructure in order: server inventory, patch management with maintenance windows, network rules and monitoring, both on-premises and in the cloud. If you are not sure which Linux servers run in your company or who maintains them, write to us or call our AI assistant Cyntia at +593 99 726 6838.
Sources
- Ubuntu Security: CVE-2026-98432 (published October 15, 2026; page with per-release status)
- Red Hat Customer Portal: CVE-2026-98432
- Bitsight: CVE-2026-98432, critical Linux TCP vulnerability
- The Hacker News: Linux Kernel TCP Flaw With 9.8 CVSS Score Affects Ubuntu, Red Hat and Other Distributions (October 2026)


