Skip to content- Sources: primary, discussion
- Summary: Jonathan Corbet's LWN report, published 2026-10-01, covers Greg Kroah-Hartman's talk at Kernel Recipes 2026 and gives kernel CVE fixes per release at about 500 before a quadrupling, which Kroah-Hartman puts at 33 CVEs fixed per day, and mean time between a bug being discovered and an exploit appearing falling from 63 days in 2018 to 7 days in 2026. Of 79 vulnerabilities a tool named Mythos reported, 24 had no detail beyond a crash, 14 were not bugs, 3 were fabricated, 15 were already fixed, and 6 of the remainder were duplicates, leaving 20 needing fixes and 10 real fixes, about one hour of work for a project merging a little over ten patches an hour. The adaptations named are documented threat-model text, security documentation that now requires maintainers be copied on reports and asks reporters to include an initial fix, and funding for a full-time developer on the kernel security contact address, with Kroah-Hartman stating the best models are still wrong 50 percent of the time and expecting a rough 18 months.
- Why it matters: The operational figure is the 7 day mean time to exploit rather than the report volume, which makes running an old kernel the exposure, and this report resolves the open question this archive left on 2026-10-03, where the same talk appeared only as a commenter paraphrase of one slide that omitted both that figure and the adaptation measures.