As a bit of an update and revising my fears:There is a chinese fork of this poc-repository which fixed an issue that delayed synack of a previous test generations syn flood would be interpreted as leak, as far as i understand it:
https://github.com/husjdodoanthing/cve-2026-98432/commit/0e5e7ffc2cb605e01bfb6b28f8d5690b262f8c23 _This version finds no issues on 6.1.190-1_What still irritates me a bit is that i can reproduce this issue on each run of the original tool on 6.1.190-1, but not once on my trixie systems with 6.12.115-2 (with identical hardware + networking parameters in sysctl - but defaults might have changed). So ultimately this might be a timing issue/network performance difference between these kernels or default parameters that came with them. -> Seems to be a false positive / bug in the poc, as a result of "late" synack retries :)
Sorry for the buzz, Robin Am 08.11.2026 um 17:12 schrieb Robin Cremer:
Dear list, quick question:Does Kernel 6.1.190-1 (build date 2026-10-16) from bookworm-security positively fix CVE-2026-98432?I updated to this kernel on one of my bookworm boxes, and initially only wanted to scan other hosts, to test for vulnerability and effectiveness of firewall rules.I used "Alucard0x1" 's poc-Project for that: https://github.com/Alucard0x1/cve-2026-98432To get a baseline of how the result for a fixed system should look, I started a local scan of this patched system and got - to my surprise - vulnerable as a result.The tool returned 10 frames that seems to contain data from other memory allocations in 180s (410000 packets tried). I obviously expected 0, which seems to be the norm for fixed kernels according to the author...[!] UAF ORACLE HIT #10 at attempt sport=63481 (expected magic 99c1d106 tag c39b8b5c) tcp 127.0.0.1:9000 -> 127.0.0.1:63481 seq=254469ff ack=d293e20a flags=0x12 SYN-ACK echoed magic d0c3b89a tag 8a99e2c0, belongs to another allocation (expected 99c1d106)ip options:0000 44 28 11 00 03 5a 21 64 03 5a 9c 67 03 5a 9c 67 D(...Z!d.Z.g.Z.g 0010 d0 c3 b8 9a 8a 99 e2 c0 00 00 00 00 00 00 00 00 ................0020 00 00 00 00 00 00 00 00 ........ frame saved to /tmp/proof.pcap (#10) uname -aLinux ***** 6.1.0-54-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.190-1 (2026-10-16) x86_64 GNU/LinuxI'm operating rather at a surface level atm, so haven't had a look at how exactly this tool works and if something else is amiss here, but packet corruption is (it being loopback) out of the question here, I guess?Would be grateful for a bit of input on this. Greetings, Robin
smime.p7s
Description: Kryptografische S/MIME-Signatur

