A newly disclosed OpenBSD kernel exploitation chain combines a shared memory accounting flaw with a separate out-of-bounds write to obtain root privileges.
Published on October 16, 2026, the research, titled “Big Fish,” includes public proof-of-concept code demonstrating exploitation despite several enabled security protections.
The oldest flaw involves a shared-memory attachment counter inherited from BSD code dating back about 37 years. That age describes the counter’s source-code lineage, not OpenBSD’s lifespan.
37-Year-Old OpenBSD Kernel Flaw
According to Calif, OpenBSD addressed the vulnerabilities through errata 026_sysvipc and 018_shmat for version 7.9. The underlying issue centers on shm_nattch, a signed 16-bit field tracking shared memory attachments.
Repeated process creation can push this counter beyond its maximum value of 32,767, wrapping it into negative territory. Incorrect accounting then enables a race condition that can produce a use-after-free vulnerability.
In the published exploitation sequence, the researcher mapped 127 System V shared memory segments, unmapped them using munmap(), and forked 259 processes.
This manipulated the attachment counter to reach −32,643. Heap grooming then used allocation statistics exposed through sysctl(KERN_MALLOCSTATS) to monitor memory placement and identify fresh allocations.
A race between segment removal through IPC_RMID and a helper process attaching another segment created the conditions needed for subsequent corruption. The exploit combined this behavior with the separate out-of-bounds write rather than relying on the counter overflow alone.
The exploit targeted an IP multicast options structure allocated through socket configuration. Corrupting its membership counters caused socket cleanup to perform atomic decrements against the kernel’s self-referential page-table mapping.
Those changes made attacker-controlled memory accessible to the supervisor, enabling further manipulation of kernel control flow.
Griffiths reported obtaining root with Kernel Address Randomized Link, Supervisor Mode Access Prevention, and Supervisor Mode Execution Prevention enabled, without requiring an information disclosure vulnerability.
The chain ultimately used linked-list removal operations to overwrite a predictable kernel pointer, execute kernel code, and restore modified state before launching a root shell.
The research highlights a distinction between randomized kernel binaries and predictable kernel allocations. Although KARL rearranges kernel code and data, Calif found that deterministic allocation behavior left certain early-boot structures at predictable locations on the same hardware.
Administrators should prioritize the relevant OpenBSD errata rather than treating existing exploitation mitigations as substitutes for patching.
This is a local privilege-escalation chain: an attacker needs the ability to execute code on the affected system, not direct remote access through this flaw alone.
Beyond the fixes, Calif proposed safe-unlinking checks, randomized page-table mappings, and less predictable kernel allocation.
These hardening proposals aim to disrupt reusable exploitation techniques; the report does not establish that every proposed defense has been adopted by OpenBSD or deployed in supported releases.
Join 16,000+ SOC teams using ANY.RUN to streamline threat investigations and reduce manual effort. Explore for your team