Exploit odds for a Linux kernel TCP bug jump to 82%

Published

A Linux kernel flaw in the code that sets up new TCP connections has gone from an obscure patch note to one of the highest-ranked bugs in the industry's exploitation forecast. CVE-2026-98432 carries a CVSS 3.1 base score of 9.8 in its National Vulnerability Database record, and on 2026-11-21 its EPSS score jumped from 0.78% to 82.1% in a single daily update, putting it in the top 0.4% of all scored CVEs.

That jump is the news. The bug itself was disclosed on 2026-10-15, and its score had sat below 1% for at least the four weeks before the jump.

A forecast, not a confirmed attack

EPSS, the Exploit Prediction Scoring System run by FIRST, estimates the probability that exploitation activity against a CVE will be observed in the next 30 days. It is a model output, and FIRST does not publish the reason behind any individual score change. An 82% reading is a strong signal to prioritise, but it is not evidence that anyone is attacking the flaw today.

As of this writing the CVE is not in CISA's Known Exploited Vulnerabilities catalog. The CVSS vector in the record rates it as reachable over the network with no privileges and no user interaction, with high impact on confidentiality, integrity and availability. The kernel developers' own description does not say what an attacker could achieve with it, and neither will we.

Two CPUs, one half-built reply

According to the kernel's fix description, the problem is a race between two processors handling the same incoming connection. When a Linux machine accepts a TCP connection, it hands the IP options stored on the half-open request over to the new child socket without copying them. If another CPU is processing a retransmitted SYN for the same request at that moment, it can still be building the SYN-ACK reply from those options when the child socket is freed. The kernel then reads memory that has already been released, a use-after-free (CWE-416) in ip_options_build().

The developers trace the exposure back to an older change that let request sockets be processed without holding the listener lock. They describe a second, similar path when request socket migration is enabled, either through net.ipv4.tcp_migrate_req=1 or a BPF_SK_REUSEPORT_SELECT_OR_MIGRATE program. The fix frees the options only after an RCU grace period, the same way other code paths already do. IPv6 is not affected by the first issue, and a related IPv6 migration case is being handled in a separate patch.

Update to 6.12.113, 6.6.159 or your branch's fix

NVD lists kernels from 4.4 onward as affected, with fixes in these stable releases:

Most organisations run distribution kernels rather than upstream stable releases, so check your vendor's advisory for the backported package version rather than matching the numbers above. The upstream fix is visible in the stable kernel tree. Linux kernel bugs have reached CISA's exploited list repeatedly this autumn, including the kernel flaw chained with PHP2Shell for root and the five kernel bugs added in three weeks in September.

Internet-facing Linux servers that accept TCP connections are the obvious place to start. A forecast this sharp, five weeks after disclosure, is the kind of signal that rewards teams who patch before the confirmation arrives rather than after.

This briefing is provided by IntelFusions for informational and defensive purposes only. It is based on sources assessed to be reliable at the time of writing, and analytic judgments carry the confidence levels indicated. Indicators of compromise are defanged; re-arm them only in controlled environments. IntelFusions is not affiliated with the organizations named and makes no warranty as to completeness or accuracy.

Earlier in Vulnerabilities

All Vulnerabilities briefings

Latest briefings

Read the full analysis on IntelFusions