AI Science Tracker

Xint Code AI scan surfaces the Linux kernel Copy Fail flaw

Linux kernel (crypto subsystem, authencesn template) · AI involvement attributed to Xint Code (Theori AI code-analysis tool)

Finding and attribution

A logic bug in the Linux kernel authencesn cryptographic template. Per the Xint account, an unprivileged local user can cause a controlled write into the page cache of readable files, with privilege escalation reported on the four distributions tested (Ubuntu 24.04, Amazon Linux 2023, RHEL 10.1, SUSE 16). The on-disk file is not modified, so on-disk checksums may miss it. Impact claims are the researchers' own and were not independently verified here.

Product: Linux kernel (crypto subsystem, authencesn template). AI system: Xint Code (Theori AI code-analysis tool). Team / organization: Theori (Xint Code Research Team; researcher Taeyang Lee).

CVE identifiers: CVE-2026-31431.

AI and human contribution (source account): Per Xint's account, Theori researcher Taeyang Lee began the work, and the finding was AI-assisted: Xint Code scanned the crypto subsystem from a short operator prompt in about an hour and ranked Copy Fail highest. Secure.com credits Lee's initial insight; Theori's press release emphasizes the tool more.

Human involvement is only described where the source account establishes it; unknown involvement does not mean none. Vendor confirmation of a flaw does not confirm every claim about AI contribution.

Discovery context: Timeline from the Xint post: reported to the kernel security team 2026-03-23, mainline patch 2026-04-01, CVE assigned 2026-04-22, public disclosure 2026-04-29. A CISA KEV listing reported by one secondary source was not verified and is not used.

Remediation status: fix documented. Patch committed to mainline 2026-04-01. Stable fixes listed by Mend as 6.18.22, 6.19.12 and 7.0 are from a secondary source and were not checked against kernel.org. Deployed patch status is not checked.

A source-documented fix does not establish installed patches or current safety. Source checking does not independently reproduce the finding or establish observed exploitation.

Disclosed 29 Apr 2026. AI attribution dated 29 Apr 2026. Sources checked 14 Nov 2026; these are separate dates, not a live installation check.

Recorded CVSS: High · highest recorded score 7.8. Documented technical effects: data modification, other, privilege escalation.

Severity, consequences and exploitation evidence

Case grouping uses the highest recorded CVSS score, with its version-specific severity band. It is descriptive, not a combined or editorial risk index. All recorded assessments remain below; scores from different versions and sources may disagree. CVSS specifications define version-specific scoring and severity bands.

  • CVE-2026-31431: CVSS 3.1 · 7.8 (High) · CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H · Assessment scenarios: AV:L - The vulnerable algif_aead path is reached through local AF_ALG socket syscalls: socket, bind, setsockopt, accept, sendmsg, and recvmsg. It is not directly reachable from remote network packets. AC:L - A local attacker can construct the AF_ALG AEAD request, choose buffer layout, AAD/tag lengths, and invoke recvmsg directly. Any race or SGL layout condition is attacker-controlled from the same process. PR:L - Creating and using AF_ALG AEAD sockets requires only an unprivileged local user on systems with CRYPTO_USER_API_AEAD enabled. The traced socket, bind, setsockopt, accept, sendmsg, and recvmsg paths do not require CAP_SYS_ADMIN or other real root privileges. UI:N - No victim interaction is needed once the attacker has local code execution. The attacker can open and drive the AF_ALG socket entirely through syscalls. S:U - The issue remains within the kernel/user crypto API security authority and does not cross a VM, IOMMU, or separate sandbox boundary. This is a standard local kernel attack surface. C:H - The removed in-place AEAD design used user-writable RX scatterlists as crypto source data and chained TX tag pages, creating a plausible decryption/authentication bypass for data protected by secret AEAD keys. Under the required uncertainty rule, this is scored as high confidentiality impact. I:H - The bug undermines AEAD authenticity by allowing source data used by the crypto operation to diverge from the data originally submitted and authenticated. This can allow forged or corrupted plaintext/ciphertext results, so integrity impact is high. A:H - The vulnerable path builds complex chained scatterlists from attacker-controlled local buffers and related AF_ALG fixes show these paths can lead to scatterwalk NULL dereferences and kernel panics. A local attacker can repeatedly trigger the operation, so availability impact is high.. Linux (CNA; CVE Program record) · checked 16 Nov 2026.
  • CVE-2026-31431: CVSS 3.1 · 7.8 (High) · CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H. redhat-SADP (ADP; CVE Program record) · checked 16 Nov 2026.

CVE-2026-31431: data modification, other, privilege escalation.
Prerequisites / affected scope: An unprivileged local user can use AF_ALG and splice to write four controlled bytes into the page cache of any readable file; the on-disk file remains unchanged. The researcher demonstrated modifying a setuid binary to obtain root on four tested distributions; the affected page cache is shared across host/container processes..
Exploitation evidence: Exploitation in the wild is not established by the cited description.. Impact source · checked 16 Nov 2026.

Documented fix: Patch committed to mainline 2026-04-01. Stable fixes listed by Mend as 6.18.22, 6.19.12 and 7.0 are from a secondary source and were not checked against kernel.org. Deployed patch status is not checked.

Original evidence

Download this source record (JSON)

Open this case in the interactive tracker · Browse all static security cases