CVE-2026-4747 is a stack-based buffer overflow in FreeBSD's RPCSEC_GSS implementation. The vulnerable code is described in the provided content as the packet-signature validation path in svc_rpc_gss_validate() within sys/rpc/rpcsec_gss/svc_rpcsec_gss.c. During validation of an RPCSEC_GSS DATA packet, the routine reconstructs RPC header material into a 128-byte stack buffer (rpchdr), writes fixed header fields, and then copies attacker-controlled credential data from the RPC authenticator into that buffer without ensuring the remaining space is sufficient. The content states that after the fixed fields are written, only 96 bytes remain, while the RPC layer may permit credential bodies up to 400 bytes, allowing an overflow of up to 304 bytes. In kernel space, this affects kgssapi.ko and is reachable through the NFS server when RPCSEC_GSS is in use. The content also states that analogous userspace RPC servers loading librpcgss_sec are vulnerable to remote code execution in the server process. The flaw does not require user interaction, and the supplied material repeatedly notes that the overflow-triggering packet validation path itself does not require prior client authentication.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
2 valid exploits after Mallory filtered fakes, detection scripts, and README-only repos.
Repository contains a README and a single Python exploit script. The README documents CVE-2026-4747 as a stack buffer overflow in FreeBSD's kgssapi.ko RPCSEC_GSS handling and explains the lab setup, exploitation strategy, and operational constraints. The main code is exploit.py, a standalone Python 3 exploit using gssapi plus raw socket/RPC helpers to attack an NFS service over RPCSEC_GSS. Core capability: the exploit performs remote kernel code execution against a vulnerable FreeBSD NFS server. It uses a 15-round attack because the overflowed credential/XDR field cannot carry the full payload in one request. Round 1 uses a ROP chain to call pmap_change_prot on kernel .bss to make it RWX. Rounds 2-14 use repeated ROP write primitives to copy shellcode into kernel memory. Final round writes the tail of the shellcode and transfers execution to it. The shellcode pivots to a controlled stack in .bss, clears DR7, calls kproc_create to spawn a kernel worker context, then uses exec_alloc_args/exec_args_add_* and kern_execve to run /bin/sh -c with a reverse-shell one-liner. The result is an interactive root shell connecting back to the attacker. The exploit is highly target-specific and operational rather than generic: it hardcodes kernel base, .bss addresses, gadget offsets, and function offsets for FreeBSD 14.4-RELEASE amd64 GENERIC with no KASLR variation. It also assumes the target exposes NFS on TCP/2049 with RPCSEC_GSS enabled and that the attacker has a valid Kerberos ticket for the configured realm/SPN. The README further notes that 2+ CPUs are effectively required because each round consumes an NFS worker thread and the exploit needs enough surviving threads to complete all 15 rounds. Repository structure is minimal and purpose-built: README.md provides vulnerability background, setup instructions for a vulnerable FreeBSD/Kerberos/NFS lab, and a walkthrough of the multi-stage ROP/shellcode design; exploit.py is the only executable entry point and contains constants, XDR/RPC helpers, shellcode generation, exploit delivery logic, listener setup, and interactive shell handling.
Repository contains a small Python exploit scaffold plus extensive Markdown documentation for CVE-2026-4747, described as a FreeBSD kernel stack buffer overflow in kgssapi.ko reachable through NFS RPCSEC_GSS authentication. Structure is: README.md with high-level exploit narrative and setup/mitigation notes, write-up.md with detailed vulnerability and exploitation methodology, claude-prompts.txt containing prompt/session history used to develop the exploit, and exploit.py as the only executable entry point. The main exploit capability is remote kernel RCE against vulnerable FreeBSD NFS servers on TCP/2049, with the intended post-exploitation outcome being a root reverse shell. The Python code defines a target profile database with hardcoded kernel addresses/gadgets for FreeBSD 14.4-RELEASE-AMD64, a simple ROP chain builder, and logic intended to generate a payload that calls pmap_change_prot on kernel BSS memory before staging shellcode. It accepts a victim IP, callback IP, and callback port. However, the implementation is incomplete: shellcode is a placeholder NOP sequence, RPC packet transmission is commented out, and there is even a likely bug using log.success(), which standard Python logging does not provide. So the repository is best characterized as an operational exploit concept/documentation set with a partial exploit scaffold rather than a turnkey weaponized tool. The write-up and README provide the substantive exploit intelligence: the vulnerable function copies attacker-controlled RPCSEC_GSS credential data into a 128-byte stack buffer without proper bounds checking, enabling overwrite of saved registers and return address in kernel context. The documented attack path requires RPCSEC_GSS/Kerberos handling to reach the vulnerable code path, and the exploit strategy uses repeated overflow packets to build/write shellcode into kernel BSS, pivot execution, invoke kernel helpers such as kproc_create, and ultimately exec /bin/sh to obtain a uid 0 reverse shell. Fingerprintable targets and artifacts include NFS on TCP/2049, Kerberos/KDC port 88 (with example forwarding to 8888), /bin/sh, /etc/krb5.conf, /boot/loader.conf, /etc/exports, and hardcoded kernel addresses for the supported target profile.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
131 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A stack overflow vulnerability in FreeBSD's NFS server that can enable remote code execution / kernel-level compromise under certain conditions, highlighted because Anthropic's model autonomously developed an exploit for it.
A remote code execution vulnerability in FreeBSD NFS described as a 17-year-old bug and highlighted as a flagship AI-assisted discovery.
A specific vulnerability presented as the strongest public exploitation case associated with Anthropic Claude Mythos Preview and Firefox-related fixes.
A previously reported FreeBSD vulnerability highlighted here because it was said to have been autonomously discovered and exploited by Mythos Preview; no technical details are provided in this content.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.