Post

FreeBSoD: Leveraging Language Models to Find and Exploit Kernel Bugs (Part 2 of 2)

Praetorian details two complete FreeBSD jail escape exploit chains combining stack and heap buffer overflows with information disclosure primitives to defeat KASLR and achieve host kernel code execution. The article covers the staged L0–L4 exploit development methodology that decomposes exploit development into isolated levels, with each level testing exactly one new exploitation primitive (KASAN trigger → control RIP → end-to-end shellcode → SMEP/SMAP bypass → full weaponization), enabling LLM assistance to produce working exploits iteratively rather than monolithically. Technical depth includes defeating stack canaries via information leaks and strategies for bypassing SMEP/SMAP to pivot userspace shellcode execution. Part 2 of a series on LLM-assisted kernel vulnerability research; builds on CVE-2026-3038 (RTSock stack overflow) and undisclosed vulnerabilities in FreeBSD networking subsystem, illustrating both the genuine utility and failure modes of LLM collaboration in offensive kernel security research.

Read original article

This post is licensed under CC BY 4.0 by the author.